KR102494953B1 - On-device real-time traffic signal control system based on deep learning - Google Patents

Abstract

A deep learning based on-device real-time traffic control system is disclosed. A traffic control method performed in an electronic device according to an embodiment includes obtaining road information related to object information from road image data for each intersection through a deep learning-based object detection and tracking model; Calculating a signal cycle for an intersection based on road information related to the obtained object information; Determining whether there is a communication connection with the server; and transmitting road information related to the obtained object information and the calculated signal period to a server when it is determined that communication with the server is connected, and road information related to the object information received from the server and A signal controller at each intersection may be controlled based on signal period information considering each intersection adjusted using the calculated signal period.

Description

Deep learning based on-device real-time traffic control system {ON-DEVICE REAL-TIME TRAFFIC SIGNAL CONTROL SYSTEM BASED ON DEEP LEARNING}

The following description relates to a technology for controlling traffic signals using deep learning technology.

The traffic signal control system is a system that analyzes the data collected by vehicle detectors installed on the road and controls the signal by setting the signal period and current time that are most suitable for the actual traffic volume. Domestic signal control systems are largely divided into general signal control, electronic signal control, and real-time signal control. Currently, real-time signal control systems generally use loop detectors to determine traffic volume. A fatal disadvantage that can occur when using a loop detector for traffic detection is that only the partial characteristics of the loop detector are utilized. This means that road conditions cannot be misjudged differently depending on where the detectors are installed, or unexpected situations such as traffic accidents cannot be dealt with.

In addition to loop detectors, there are magnetic detectors and image detectors. The problems of video detectors currently used for traffic analysis include performance deviation due to external variables such as weather and low performance compared to other detectors. Recently, the performance of computer vision using deep learning has risen dramatically and has consistently high performance in various situations. As a result, image detectors using deep learning models can compensate for the shortcomings of existing image detectors. However, unlike previous computer vision models, deep learning models basically require a high-performance server because they require a large amount of computation, resulting in excessive costs and the need to transmit an image containing a license plate, one of important personal information, to the server. Privacy issues exist.

It is possible to provide a real-time traffic signal control system and method for controlling a signal optimized for road conditions in real time by grasping a deep learning model of traffic volume on-device.

A traffic control method performed in an electronic device includes obtaining road information related to object information from road image data of an intersection through a deep learning-based object detection and tracking model; Calculating a signal cycle for an intersection based on road information related to the obtained object information; Determining whether there is a communication connection with the server; and transmitting road information related to the obtained object information and the calculated signal period to a server when it is determined that communication with the server is connected, and in the traffic control method, the object transmitted from the server A signal controller at each intersection can be controlled based on road information related to information and signal period information considering each intersection adjusted using the calculated signal period.

The acquiring step may include collecting road image data of an intersection captured by a photographing device installed at each intersection, inputting the collected road image data of the intersection into the deep learning-based object detection and tracking model, and Extracting road information related to object information from the collected image data of the intersection using a deep learning-based object detection and tracking model, and de-identifying the road information related to the extracted object information. can

The acquiring may include recognizing types of vehicles from the collected road image data for the intersection, tracking position changes of the recognized vehicles, and passing the recognized vehicles per unit time on the road including each intersection. and extracting road information related to object information including a traffic volume indicating a traffic volume and a degree of saturation indicating a share occupied by the recognized vehicles on the road including each intersection.

The calculating may include calculating the present time based on the traffic volume of the road including the intersection and the traffic volume of the road including the intersection per unit time through road information related to the obtained object information, and calculating the present time of the road at the calculated present time. and calculating the present time for the intersection through a process of comparing the information with road information of the present time before the calculated present time and assigning weights and subtraction values to roads including the intersection.

In the calculating step, if there are other vehicles after the calculated presenting time based on the calculated presenting time based on the degree of saturation of the road including the intersection, a weight is assigned to the vehicle within the calculated presenting time. If fields are empty, a subtraction value is given, and the calculated road information at present time is compared with the road information at present time before the calculated present time, and a weight and a subtraction value are assigned to roads including intersections. It may include calculating the present time for.

The determining step includes operating the signal controller to control traffic lights based on the calculated signal cycle through communication between the electronic device and the signal controller when it is determined that communication with the server will not be connected. The signal controller may receive the calculated signal period from the electronic device and operate the traffic light through the received signal period.

The server collects road information about the minimum control unit area composed of intersections, calculates the number of vehicles entering and exiting each intersection through the number of vehicles going straight, turning left, and turning right, and considering the entering and leaving vehicles. Thus, the weight is reflected in the present time of the adjacent intersection adjacent to the intersection, and the offset time and signal period information of each intersection can be calculated based on the intersection.

In the server, when the number of vehicles departing from the adjacent intersection in the direction of the intersection exceeds a preset standard, a weight is given to the present time of road information including the intersection, and the traffic volume of the adjacent intersection, the amount of entry into the intersection, and saturation The offset time may be adjusted in consideration of

The server may reduce the offset time when the amount of entry and saturation of the intersection are equal to or greater than a preset reference, and increase the offset time when the amount of entry and saturation of the intersection are less than or equal to a preset reference.

An electronic device for traffic control includes: an acquisition unit that obtains road information related to object information from road image data for each intersection through a deep learning-based object detection and tracking model; a signal calculator calculating a signal period for an intersection based on road information related to the obtained object information; a communication determination unit that determines whether or not there is a communication connection with the server; and a data transmitter for transmitting road information related to the obtained object information and the calculated signal period to a server when it is determined that communication with the server is connected, and road information related to the received object information from the server. The signal controller of each intersection can be controlled based on the signal period information considering each intersection adjusted using the calculated signal period and the signal period.

Privacy of each vehicle can be protected by controlling the signal cycle using road information related to de-identified object information.

Even when the network is disconnected, the signal controller at the intersection can be independently controlled by calculating the signal period for the intersection through deep learning operation at the edge terminal.

By calculating the signal period for each intersection in the server using the road information and signal period transmitted from the edge terminal, the signal period considering the influence of the adjacent intersection can be operated.

After the edge terminal calculates the signal period for the intersection, the server can control and operate traffic signals more accurately and quickly by considering and optimizing the signal period for each intersection with the adjacent intersection.

1 is a diagram for explaining the configuration of a traffic control system according to an embodiment.
2 is a diagram for explaining a traffic control operation in a traffic control system according to an embodiment.
3 is a diagram for explaining the configuration of a minimum control unit according to an embodiment.
4 is a block diagram for explaining the configuration of an edge terminal according to an embodiment.
5 is a flowchart illustrating a traffic control method in an edge terminal according to an embodiment.
6 is an example for explaining an operation of obtaining road information related to object information through a deep learning-based object detection and tracking model, according to an embodiment.
7 is a flowchart illustrating an operation of controlling signal period information in a server according to an exemplary embodiment.
8 is an example for describing an object detection model according to an embodiment.
9 and 10 are examples for explaining a tracking model 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 traffic control system according to an embodiment.

The traffic control system is for controlling on-device real-time traffic signals based on deep learning, and may include an electronic device 100, a server 110, a signal controller 120, and a photographing device 130. The traffic control system may be composed of a field unit and a server.

For example, the electronic device 100, the signal controller 120, and the photographing device 130 may be installed and operated in a field. The site unit may include devices installed at road sites including each intersection.

The electronic device 100 may be a device for collecting and analyzing traffic information in the field. In this case, the electronic device 100 may analyze traffic information about the intersection using data related to the road including the intersection. The electronic device 100 may transmit and receive data with the signal controller 120 and the photographing device 130 through wired/wireless communication. Hereinafter, the electronic device 100 will be described as an edge device.

The edge terminal 100 may receive road image data of an intersection through the photographing device 130, and use a deep learning-based object detection and tracking model from the received road image data of an intersection to obtain road information related to object information. can be extracted. The edge terminal 100 may de-identify data related to personal information from road information related to the extracted object information. The edge terminal 100 may transmit de-identified data (road information related to object information) to the server 110 . At this time, the edge terminal 100 collects and analyzes the road image data received from the photographing device 130 in the field, and then transmits the data to the server 100, so the server 100 transmits the data for each intersection. It is possible to control signal information faster than collecting and analyzing each. In addition, it is possible to prevent data transmitted/received from the server 100 from being interrupted in the middle or performing analysis using incorrect data.

The server 110 may exist in a location other than the site and receive data transmitted from the edge terminal 100 to generate an optimized signal cycle for each intersection. The server 110 may transmit the generated signal period to each signal controller 120.

The signal controller 120 may mean a device that controls the signal period of traffic lights related to each intersection. Traffic lights may be operated based on the signal period controlled by the signal controller 120 .

The photographing device 130 may photograph a road including each intersection. In this case, the photographing device 130 may be installed at a specific location to photograph a road including an intersection. The photographing device 130 may correspond to at least one camera for photographing a road including an intersection, 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.

In the embodiment, the edge terminal 100, the signal controller 120, and the photographing device 130 are configured for each intersection, and the road including each intersection is photographed and analyzed, and the analyzed signal period is transmitted to the signal controller 120. It can be. Accordingly, the signal controller 120 may reflect the signal period to the traffic light associated with the intersection.

According to one embodiment, by drastically reducing the amount of calculation without deteriorating the performance of the deep learning model through lightweight technology, it is possible to operate on-device in a mobile or SBC (Single Board Computer) rather than a high-performance server. . By using this, it is possible to solve the cost problem and the personal information problem because the on-device de-identified road information is transmitted to the server.

2 is a diagram for explaining a traffic control operation in a traffic control system according to an embodiment.

In FIG. 2, the CCTV 130 will be described as an example of a photographing device. Road image data may be acquired as a road including an intersection is photographed by the CCTV 130 . Road image data photographed by the CCTV 130 in real time may be collected, and the captured road image data may be accumulated for a predetermined period. Road image data collected by the CCTV 130 may be transmitted to the edge terminal 100 . For example, road image data collected by the CCTV 130 may be transmitted in real time or periodically/non-periodically.

The edge terminal 100 may receive road image data from the CCTV 130 . The edge terminal 100 may recognize object information from road image data using a deep learning-based object detection and tracking model. 6, 8, 9, and 10 are examples for explaining an operation of obtaining road information related to object information through a deep learning-based object detection and tracking model. Road image data 601 may be input to the deep learning-based object detection and tracking model 600 . The deep learning-based object detection and tracking model 600 refers to a model that detects an object from the road image data 601 and tracks a positional change of the detected object. The deep learning-based object detection and tracking model 600 is a model configured for object detection and tracking, and may be trained using a data set for object detection and tracking. In an embodiment, road image data 601 may be input to the deep learning-based object detection and tracking model 600 trained using the data set. In this way, road information 602 related to object information may be obtained from road image data 601 using the trained deep learning-based object detection and tracking model 600 . In this case, the road information 602 related to the object information may include object information detected from the road image data 601 and road information including the object information (eg, traffic volume). For example, the type of vehicle (e.g., car, truck, ambulance, forklift, motorcycle, or general vehicle such as car, bus, truck, special vehicle such as ambulance, police car, fire engine, emergency of special vehicle (warning light) turned on) can be recognized and tracked. Also, roads can be recognized. In addition, the amount of traffic passing by vehicles per unit time on the road, saturation indicating the occupancy of the road by vehicles, and the like may be obtained as road information 602 related to object information. Traffic volume refers to the number of vehicles passing through a defined Region of Interest (ROI) per unit time. Saturation means (the number of vehicles present in a specific range) / (the number of vehicles that can be accommodated in the range). In this case, the number of vehicles existing within a specific range means the number of vehicles recognized by a camera in a continuous road entering an ROI area defined by one or more cameras.

In detail, the deep learning-based object detection and tracking model 600 may detect an object from road image data and obtain object information by tracking a positional change of the object. For example, the deep learning-based object detection and tracking model 600 may consist of an object detection model and a tracking model, respectively, or may be configured in a combined form. In addition, it is configured in various forms to detect an object through a deep learning-based object detection and tracking model 600, and the detected object can be tracked.

First of all, the object detection model will be described. 8 is an example for describing an object detection model according to an embodiment. Recently, the performance of computer vision has risen by leaps and bounds due to the rapid growth of artificial intelligence and deep learning. The object detection model is largely divided into a 1-stage based model and a 2-stage based model. It refers to a model that has the ability to distinguish the location and type of an object while learning rules within unstructured data where the probability distribution is difficult to define, such as images.

The 1-stage based model is a method of performing localization to find the location of an object and classification to classify what the object is at once. Representative models such as YOLO and SSD exist. It has a faster speed with a more efficient structure than the 2-stage based model, but has relatively low performance.

The 2-stage based model first proceeds with the region proposal step in the localization to find the location of the object, and then proceeds with the classification step to explore the type of object. Representatively, R-CNN model exists. Although it shows higher performance than the 1-stage based model, it has the disadvantage of being slower due to its more complex structure.

9 and 10 are examples for explaining a tracking model according to one embodiment. The object tracking model is a field that has been studied for a long time in computer vision and image processing, and is a field that tracks the positional change of a specific object in a video image. In the existing object tracking field, there were methods such as Point Tracking and Kernel Tracking, but with the recent rapid growth of deep learning, models using deep learning are being actively studied. In the field of object tracking using deep learning, there are correlation filter based and Siamese Network based.

Referring to FIG. 9 , it is an example illustrating an object tracking operation based on a correlation filter. In the correlation filter-based field, it is a method of tracking an object by finding a part that has a high correlation with a target object using features extracted by a deep learning model.

Referring to FIG. 10, it is an example of an object tracking operation based on a Siamese network. A Siamese network is a network that compares the similarity between two images in the field of deep learning. When features are extracted with a convolution network, similar images learn to have a close feature distance, and a network that learns as the distance between features extracted from other images increases. am.

The edge terminal 100 may recognize object information from road image data using a deep learning-based object detection and tracking model. The edge terminal 100 may obtain road information related to object information. The edge terminal 100 may perform de-identification of data related to personal information such as a license plate and a vehicle model. For example, the edge terminal 100 may perform special processing such as mosaic and blur on personal information-related data among road information related to object information, and may encrypt personal information-related data. In this way, de-identification may proceed through special processing such as mosaic, blur, and encryption. In other words, with respect to object information and related road information, some data may be de-identified in a closed form.

The edge terminal 100 may calculate a signal cycle optimized for an intersection based on road information related to object information. The edge terminal 100 may determine the traffic volume of roads including intersections per unit time. For example, vehicle traffic per 15 seconds can be grasped. The edge terminal 100 may calculate the present time based on the traffic volume of the road including the intersection. The edge terminal 100 may assign a weight to the present time of the road (intersection) through the degree of saturation in case of a road (intersection) where vehicles remain even after the present time. The edge terminal 100 may give a subtraction value to the present time of the road through the degree of saturation when all vehicles have passed within the present time and the road is empty. The edge terminal 100 compares the present time with road information (eg, traffic volume, saturation) of the previous present time, and can assign weights and subtraction values to the roads within the intersection by identifying the increase and decrease of the state tax. The edge terminal 100 may calculate an optimized present time for an intersection using the calculated weight and subtraction value for the existing present time. Accordingly, the edge terminal 100 may calculate the signal period for the intersection based on the present time.

Existing signal information of each intersection can be grasped. For example, the existing period is 150 seconds, and the existing present time may be composed of 70 seconds, 25 seconds, 25 seconds, and 30 seconds, respectively. In addition, minimum and maximum times including 1 prefecture (30, 105), 2 prefecture (130, 35), 3 prefecture (13, 40), and 4 prefecture (15, 50) may be set for each prefecture. . The number of passing vehicles per unit time of the intersection can be calculated. If the unit time is 15 seconds and the traffic volume per unit time is 5 vehicles, the traffic volume per unit time of the three-lane road can be derived as 15 vehicles through 5*3. Saturation of the intersection can be calculated. When the maximum number of detected vehicles in the ROI is 15 and the number of detected vehicles is 5, saturation can be derived as 5/15 = 33%. Present time can be calculated by considering traffic volume and saturation in the existing signal information. Traffic volume for each prefecture can be calculated (eg, 30 prefectures in 1 prefecture, 15 vehicles in 2 prefectures, 15 vehicles in 3 prefectures, 25 vehicles in 4 prefectures). In addition, the degree of saturation for each display may be calculated (eg, 50% for 1 display, 0% for 2 display, 10% for 3 display, 30% for 4 display). In addition, the priority for each prefecture may be set based on traffic volume (eg, 1st prefecture, 2nd prefecture, 3rd prefecture, 3rd prefecture, 4th prefecture, 2nd priority). In addition, a weight value for each present time may be set based on the priority. Each display time can be adjusted based on saturation. For example, the display time of roads (intersections) with saturation of 50% to 100% or more is +10 seconds, the display time of roads (intersections) with saturation of 0% to 50% or more is +5 seconds, and roads with saturation of 0% (Intersection) display time can be adjusted to -5 seconds. Accordingly, the existing display times are 70, 25, 25, and 30, and the new display times 82, 20, 26, and 37, and the new period can be adjusted from 150 seconds to 165. The above-mentioned operation may be repeatedly performed, and new current time and period information may be transmitted to the server.

The edge terminal 100 may transmit road information and calculated signal information related to the de-identified object information to the server 110 . In addition, the edge terminal 100 may transmit the calculated signal period to the signal controller 120 . When communication with the server is not smooth due to a network problem, the edge terminal 100 may operate the signal controller 120 through communication between the edge terminal 100 and the signal controller 120 .

The server 110 may receive road information related to de-identified object information and the calculated signal period from the edge terminal 100 . The server 110 may calculate a signal period for each intersection using road information related to the de-identified object information and the calculated signal period. In detail, the server 110 may optimize signal period information for each intersection by receiving road information related to de-identified object information for each intersection and the calculated signal period.

Referring to FIG. 7 , in step 710, the server 110 may calculate the offset time and signal period information of each intersection by collecting road information and signal period related to object information on a plurality of intersections. The server 110 may collect intersection road information of a minimum control unit area, which is a group composed of intersections having similar traffic patterns. The server 110 may calculate the number of vehicles entering and exiting each intersection through the number of vehicles going straight, turning left, and turning right in order to consider the influence of adjacent roads. The server 110 may reflect a weight to the present time of an adjacent intersection in consideration of an entering vehicle and an leaving vehicle. The server 110 may calculate offset time and signal period information of each intersection based on the important intersection in the minimum control unit. The server 110 may assign a weight to the current time of the corresponding intersection if there are more deviating vehicles than a preset standard in the direction of a specific intersection at an adjacent intersection.

In step 720, the server 110 may adjust the offset time according to the traffic volume of each intersection, the amount of entry of an intersection adjacent to each intersection, and the degree of saturation. The server 110 may adjust the offset time so that entering vehicles can smoothly pass through the intersection without stopping by reducing the offset time if the entry amount and saturation are large, and increasing the offset time if the entry amount and saturation are small.

In step 730, the server 110 may transmit signal period information considering the influence of each intersection to each signal controller 120 through adjustment. The server 110 may transmit the calculated signal period information to the signal controller 120 for each intersection. The signal controller 120 may receive a signal period from the edge terminal 100 or the server 110 . The signal controller 120 may reflect the received signal cycle to traffic lights. According to an embodiment, since information on each intersection is transmitted to the server because it is affected by the adjacent intersection, it is possible to operate a signal cycle considering the influence of the adjacent intersection.

For example, the server 110 may adjust the present time considering an adjacent intersection. New present time and cycle information for each intersection may be received. For example, the display time for intersection A (70, 25, 25, 30) has been adjusted to the new display time (80, 20, 26, 37), and period information that was 150 can be adjusted to 165, and at intersection B It can be seen that the present times (40, 35, 45, 25) for the new present times (65, 40, 55, 40) have been adjusted. The server 110 may adjust present time and cycle information associated between adjacent intersections. For example, the server 110 adjusts the presenting time (82, 20, 26, 37 -> 85, 21, 28, 42) and period information (165 -> 176) for the intersection A, and the presenting time for the intersection B It can be adjusted by time (65, 40, 55, 40 -> 62, 36, 48, 35) and period information (200 -> 181). The server 110 may transmit the current time considering the adjacent intersection to each intersection. The server 110 may transmit the present time for each intersection to each signal controller 120 . For example, the server 110 transmits the final presenting times (85, 21, 28, 42) and cycle information 176 of the intersection A to the signal controller 120 for the intersection A, and transmits the signal controller 120 for the intersection B. (120) can transmit the final appearance time (62, 36, 48, 35) and cycle information (181) of the B intersection.

3 is a diagram for explaining the configuration of a minimum control unit according to an embodiment.

The current real-time signal controller has a disadvantage in that it needs to be adjusted by a person at the control center or to send repetitive signal cycles for each time zone based on a Time of Day (TOD) plan. In the embodiment, in order to solve this problem, optimized signal control is enabled by reflecting road information in real time.

The minimum control unit may mean that the National Police Agency, Road Traffic Authority, and the like identify geographical characteristics and group at least one intersection. Depending on geographical characteristics, at least one intersection in a specific area may be grouped as a minimum control unit. For example, as you move out of the city center, you may be less affected by vehicles. Accordingly, it can be classified into important intersections that are heavily influenced by traffic, semi-important intersections that are less influenced by traffic, and non-important intersections that are hardly affected by traffic.

In addition, the present time will be described. Display time means providing vehicles at an intersection so that they can pass as a whole, and signal commands such as going straight, turning left, turning right, etc. based on a specific location (road direction at a specific location) are called display. For example, if the first appearance is to go straight, the second appearance is to turn left, and the third appearance is to turn right, it is called a signal cycle when each appearance is all progressed.

4 is a block diagram illustrating a configuration of an edge terminal according to an embodiment, and FIG. 5 is a flowchart illustrating a traffic control method in an edge terminal according to an embodiment.

The processor of the edge terminal 100 may include an acquisition unit 410, a period calculation unit 420, a communication determination unit 430, and a data transmission unit 440. Components of the processor may represent different functions performed by the processor according to control commands provided by program codes stored in the vehicle control device. The processor and components of the processor may control the edge terminal to perform steps 510 to 541 included in the traffic control method of FIG. 5 . In this case, the processor and components of the processor may be implemented to execute instructions according to the code of an operating system included in the memory and the code of at least one program.

The processor may load a program code stored in a file of a program for a traffic control method 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 acquisition unit 410, the period calculation unit 420, the communication determination unit 430, and the data transmission unit 440 executes a command of a corresponding part of the program code loaded into the memory to perform subsequent steps (510 to 510). 541) may be different functional representations of the processor.

In step 510, the acquisition unit 410 may obtain road information related to object information from road image data for each intersection through a deep learning-based object detection and tracking model. The acquisition unit 410 collects road image data for the intersection captured by a photographing device installed at each intersection, inputs the collected road image data for the intersection to the deep learning-based object detection and tracking model, and deep learning It is possible to extract road information related to object information from the image data of the intersection collected using the based object detection and tracking model, and to de-identify the road information related to the extracted object information. The acquisition unit 410 recognizes the type of vehicles from the collected road image data of the intersection, tracks the change in the location of the recognized vehicles, and recognizes the amount of traffic that the recognized vehicles pass per unit time on the road including each intersection. It is possible to extract road information related to object information including saturation representing the occupancy rate of vehicles on the road including each intersection.

In step 520, the cycle calculation unit 420 may calculate a signal cycle for an intersection based on road information related to the acquired object information. The period calculation unit 420 calculates the present time based on the traffic volume of the road including the intersection and the traffic volume of the road including the intersection per unit time through road information related to the obtained object information, and calculates the current time based on the road information of the calculated present time. The present time for the intersection can be calculated through the process of assigning weights and subtraction values to the roads including the intersection by comparing the calculated present time with the road information of the present time before the calculated present time. The period calculation unit 420 assigns a weight to vehicles remaining empty after the calculated present time based on the present time calculated through the degree of saturation of the road including the intersection, and assigns a weight to vehicles within the calculated present time. If there is, a subtraction value is given, and the road information of the calculated present time is compared with the road information of the present time before the calculated present time, and the present time for the intersection is given a weight and a subtraction value to the road including the intersection. can be calculated.

In step 530, the communication determination unit 430 may determine whether there is a communication connection with the server. For example, the communication determining unit 430 may transmit the calculated signal cycle for the intersection to the server when determining whether or not there is a communication connection. If the signal period for the calculated intersection is transmitted to the server, it can be determined that communication with the server is connected, and if the signal period for the calculated intersection is returned without being transmitted to the server, it is assumed that communication with the server is not connected. can judge Alternatively, the communication determination unit 430 may check the communication status to determine whether there is a communication connection with the server.

In step 540, the data transmission unit 440 may transmit road information related to the acquired object information and the calculated signal period to the server when it is determined that communication with the server is connected. Therefore, the signal controller of each intersection can be controlled based on the signal period information considering each intersection calculated using the road information related to the object information transmitted from the server and the calculated signal period. In step 541, the data transmission unit 440 may operate the signal controller so that traffic lights are controlled based on the signal cycle calculated through communication between the electronic device (edge terminal) and the signal controller.

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. You can command the device. 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 (10)

In the traffic control method performed in the electronic device,
obtaining road information related to object information from road image data of an intersection through a deep learning-based object detection and tracking model;
calculating a signal cycle for the intersection based on road information related to the acquired object information;
transmitting the calculated signal period to a server;
when the signal period is returned from the server, the electronic device transmitting the calculated signal period to the signal controller of the intersection so that traffic lights are controlled based on the calculated signal period; and
If the signal period is not returned from the server, the electronic device determines that communication with the server is established, and further transmits road information related to the acquired object information to the server.
Traffic control method comprising a. According to claim 1,
The obtaining step is
Road image data for the intersection captured by a photographing device installed at the intersection is collected, the collected road image data for the intersection is input to the deep learning-based object detection and tracking model, and the deep learning-based object detection and extracting road information related to object information from the collected image data of the intersection using a tracking model, and de-identifying the road information related to the extracted object information.
Traffic control method comprising a. According to claim 2,
The obtaining step is
Recognizing the types of vehicles from the collected road image data of the intersection, tracking the position change of the recognized vehicles, and the traffic volume that the recognized vehicles pass per unit time on the road including the intersection, the recognized vehicle extracting road information related to object information including saturation indicating occupancy rates occupied by the roads including the intersection;
Traffic control method comprising a. According to claim 1,
The calculating step is
Based on the road information related to the acquired object information, the present time is calculated based on the traffic volume of the road including the intersection per unit time and the traffic volume of the road including the intersection, and the road information of the calculated present time and the calculation Calculating the present time for the intersection through a process of assigning a weight and a subtraction value to the road including the intersection compared with road information of the present time before the present time.
Traffic control method comprising a. According to claim 4,
The calculating step is
Based on the calculated presenting time through the degree of saturation of the road including the intersection, if there are remaining vehicles after the calculated presenting time, a weight is given, and if vehicles are empty within the calculated presenting time, Through the process of assigning a subtraction value and comparing the calculated road information of the present time with the road information of the present time prior to the calculated present time, a weight and a subtraction value are assigned to the road including the intersection. steps to calculate time
Traffic control method comprising a. delete delete delete delete In electronic devices for traffic control,
an acquisition unit that obtains road information related to object information from road image data for each intersection through a deep learning-based object detection and tracking model;
a signal calculating unit calculating a signal period for the intersection based on road information related to the acquired object information;
a communication determination unit transmitting the calculated signal period to a server; and
When the signal period is returned from the server, the calculated signal period is transmitted to the signal controller of the intersection so that traffic lights are controlled based on the calculated signal period, and when the signal period is not returned from the server A data transmitter for determining that communication with the server is established and further transmitting road information related to the acquired object information to the server
including,
Electronic device characterized in that.

Images