KR102604816B1 - AI camera apparatus preventing vehicle traffic accident and controlling method thereof - Google Patents

Abstract

The present invention relates to an AI camera device and a control method for preventing vehicle traffic accidents, including a GPS module that receives GPS signals and outputs location information of the vehicle, and a camera that photographs one side of the vehicle and outputs an image of one side of the vehicle. A module, a pattern storage unit that stores a segment pattern image divided into a plurality of segment areas, and a device that receives the image of one side of the vehicle and the segment pattern image respectively and outputs an overlapping image in which the segment pattern image is superimposed on the image of one side of the vehicle. If the image overlap part and the vehicle moving speed are within a preset range, the overlapping image is analyzed and there is an object in multiple partitions of the segment pattern image, and if the object is a moving object, the speaker is turned on or off according to the segment area where the moving object is located. It is characterized by including a controller that controls and outputs an alarm sound output through a speaker by setting the signal cycle differently.

Description

AI camera apparatus preventing vehicle traffic accident and controlling method thereof}

The present invention relates to an AI camera device and a control method for preventing vehicle traffic accidents. In particular, the camera is placed on the upper side of a high-height vehicle to divide a wide area, including blind spots, on one side or the other side of the vehicle into multiple divided areas. After dividing the images into multiple segments and checking whether there are moving objects in multiple divisions captured according to the moving speed of the vehicle, if there are moving objects, different warnings are issued depending on the divisions where the moving objects are located to prevent vehicle traffic. This relates to an AI camera device and its control method for preventing traffic accidents in vehicles that can prevent accidents.

Vehicles must accurately recognize traffic information to safely protect people walking on the road. Traffic information must be accurately transmitted to vehicles, and technologies are being developed to prevent traffic accidents caused by vehicles by transmitting traffic information to vehicles, and related technologies are disclosed in Korean Patent Publication No. 10-2305577 (Patent Document 1). It is done.

Patent Document 1 relates to a system to prevent traffic accidents caused by vehicles, and includes a rumble strip, a right turn indicator lamp, a traffic barrier, a warning lamp, and a controller.

The rumble strip is installed on the road just before a right turn to cause driving vibration as a warning, and the indicator lamp for whether a right turn is possible is installed in front of the rumble strip on the road just before a right turn and emits light from the LED buried in the road surface depending on whether the pedestrian traffic light shows a walking signal. Indicates whether a right turn is possible. The traffic barrier is installed on the entrance side of the crosswalk on the side of the road immediately after a right turn and operates depending on whether the pedestrian signal is a walking signal. The warning lamp is installed on the top of a bollard or fence installed on the side of the road just before a right turn and is a pedestrian signal light. Lights up when the walk signal is given. The controller controls the operation of the right turn indicator lamp, traffic blocker, and warning lamp depending on whether the pedestrian traffic signal is a walking signal or not.

A vehicle traffic accident prevention device, such as Patent Document 1, warns of vehicles turning right when turning right through a rumble strip, and the controller controls the operation of the right turn indicator lamp, traffic blocker, and warning lamp according to the walking signal of the pedestrian traffic light to prevent the vehicle from moving. By deciding whether to turn right, people walking on the road are protected.

Conventional vehicle traffic accident prevention devices, such as those in Patent Document 1, require separate equipment such as traffic barriers or warning lamps on the road, so they are not easy to apply to all roads or road intersections, and are not easy to apply to vehicles such as route buses with high overall height. Since blind spots are wide, there is a problem that traffic accidents may occur due to blind spots even when separate facilities are provided to prevent vehicle accidents.

: Korean Patent Publication No. 10-2305577

The purpose of the present invention is to solve the above-mentioned problem, by placing a camera on the upper side of a vehicle with a high overall height, dividing one side or the other side of the vehicle into a large area, including blind spots, into a plurality of divided areas, and taking pictures while the vehicle moves. Vehicle traffic accidents can be prevented by checking whether there are moving objects in multiple divisions captured according to speed and then issuing different warnings depending on the divisions where the moving objects are located if there are moving objects. The goal is to provide an AI camera device and its control method to prevent accidents.

The AI camera device for preventing traffic accidents of vehicles of the present invention includes a GPS module that receives GPS (Global Positioning System) signals and outputs location information of the vehicle, a camera module that photographs one side of the vehicle and outputs an image of one side of the vehicle, and , a pattern storage unit that stores a segment pattern image divided into a plurality of segment areas, and a pattern storage unit that receives the image of one side of the vehicle and the segment pattern image, respectively, and outputs an overlapping image in which the segment pattern image is superimposed on the image of one side of the vehicle. An image overlapping unit receives the overlapping image and transmits it to the display to display the overlapping image, receives the location information of the vehicle, calculates the moving speed of the vehicle, and analyzes the overlapping image if the moving speed of the vehicle is within a preset range. If there is an object in multiple partitions of the segment pattern image and the object is a moving object, the cycle of the speaker on-off signal is set differently depending on the segment where the moving object is located, and an alarm is output through the speaker. It is characterized by including a controller that controls and outputs sound.

The AI camera control method for preventing traffic accidents of vehicles of the present invention includes calculating the vehicle's moving speed using the vehicle's location information, and once the vehicle's moving speed is calculated, checking whether the vehicle's moving speed is within a set range. Step, if the moving speed of the vehicle is within a set range, checking whether there is an object in the overlapping image in which a segment pattern image including a plurality of segment areas is overlapped with the image of one side of the vehicle, and if the object is in the plurality of segment areas A step of checking where the object is located and whether the object is a moving object or a fixed object using a machine learning method, and if the object is a moving object, the alarm generation speed is adjusted according to the partition area in which the object is located. It includes a step of determining, and in the step of checking using the machine learning method, the machine learning method is Deep Learning or Bayesian Ridge, and in the step of determining the alarm occurrence rate, the alarm The generation speed is characterized as the generation period of the speaker on-off signal.

The AI camera device and its control method for preventing traffic accidents in vehicles of the present invention places a camera on the upper side of a vehicle with a high overall height, divides one side or the other side of the vehicle into a large area, including blind spots, and captures the image. , it is possible to prevent vehicle traffic accidents by checking whether there is a moving object in multiple partitions captured according to the moving speed of the vehicle, and then issuing different warnings depending on the partition where the moving object is located if there is a moving object. There is an advantage.

1 is a perspective view of a vehicle equipped with an embodiment of an AI camera device for preventing traffic accidents in vehicles of the present invention;
Figure 2 is a block diagram of the AI camera device for preventing traffic accidents in vehicles of the present invention shown in Figure 1;
Figure 3 is a diagram showing a first embodiment of the segmented area pattern image shown in Figure 1;
Figure 4 is a diagram showing a second embodiment of the segmented area pattern image shown in Figure 3;
Figure 5 is a diagram showing a third embodiment of the segmented area pattern image shown in Figure 3;
Figure 6 is a perspective view of a vehicle showing another embodiment of the AI camera device for preventing traffic accidents of the vehicle in Figure 1;
FIG. 7 is a flowchart showing a control method for preventing traffic accidents of vehicles using the AI camera device shown in FIG. 2.

Hereinafter, an embodiment of the AI camera device and its control method for preventing traffic accidents of vehicles of the present invention will be described with reference to the attached drawings.

As shown in Figures 1 and 2, the AI camera device 100 for preventing traffic accidents of vehicles of the present invention includes a GPS (Global Positioning System) receiver 110, a camera module 120, a pattern storage unit 130, and an image It is configured to include an overlapping unit 140 and a controller 150.

The GPS module 110 receives a GPS signal and outputs location information of the vehicle 200, and the camera module 120 photographs one side of the vehicle 200 and outputs an image of one side of the vehicle. The pattern storage unit 130 stores a segment pattern image (PP1) divided into a plurality of segment areas (PA1, PA2,..., PAn: where n is a natural number), and the image overlapping unit 140 stores the camera module ( The image of one side of the vehicle and the segment pattern image PP1 output from the pattern storage unit 130 are respectively received from 120) and output an overlapping image in which the segment pattern image PP1 is superimposed on the image of one side of the vehicle.

Control the image overlapping unit 140 to receive the overlapping image and transmit it to the display 151 to display the overlapping image, and receive the location information of the vehicle 200 to calculate the moving speed of the vehicle 200 to move the vehicle. If the speed is in a preset range, the overlapping image is analyzed and there is an object in multiple segment areas (PA1, PA2,...,PAn) of the segment pattern image (PP1), and if the object is a moving object, the segment area where the moving object is located Accordingly, the cycle of the speaker on-off signal is set differently to control and output the alarm sound output through the speaker 152.

A specific embodiment of the AI camera device 100 for preventing traffic accidents of vehicles according to the present invention will be described as follows.

The AI camera device 100 for preventing traffic accidents of vehicles of the present invention includes a GPS module 110, a pattern storage unit 130, an image overlapping unit 140, and a controller 150 disposed inside the case 110b. And the camera module 120 is disposed in the case 110b connected to the bracket 100a so as to protrude outward as shown in FIG. 1. The bracket 100a is placed on the upper side of one side of the vehicle 200, and the case 110b is connected to and supports both ends. Here, the vehicle 200 is applied to have a height of 2.0 to 4.0 m (meters) and a length of 10 to 13 m from the ground where the wheels 200a are in contact. That is, the vehicle 200 is applied to a large bus or a large truck, such as a high route or express bus, with a height of 2.0 to 4.0 m above the ground where the full height wheel 200a is in contact.

As shown in FIG. 2, the GPS module 110 receives GPS signals transmitted from a Global Positioning System (GPS) satellite (not shown) in real time and outputs location information of the vehicle 200.

As shown in FIGS. 1 and 2, the camera module 120 is disposed to protrude on one side of the case 110b of the AI camera device 100 for preventing traffic accidents of the vehicle of the present invention, so as to project one side of the vehicle 200. Take a picture and output an image of one side of the vehicle. The image of one side of the vehicle includes a blind spot that occurs as a high characteristic in the vertical direction (Z) of the vehicle 200 because the vehicle 200 is a large bus or truck.

The camera module 120 is used with a horizontal viewing angle of 150 to 170 degrees, and is installed on one side of the vehicle 200 at a height (HT) of 2.5 to 3.5 meters above the ground where the wheels 200a are in contact. The segmented area pattern image (PP1) is arranged to have a width (WP) of at least 2.5 to 3.5 m and a length (LP) of 5.5 to 6.5 m, and a characteristic that is high in the vertical direction (Z) of the vehicle 200, that is, has a high overall height. The camera captures the wide blind spot caused by the camera and outputs a one-sided image. That is, when the camera module 120 is arranged so that the horizontal viewing angle is 170° and the height (HT) is 3 m on one side of the vehicle 200 with a high overall height, the camera module 120 covers the blind spot on one side of the vehicle 200 as shown in FIG. 1. Including, the width (WP)

The pattern storage unit 130 uses a non-volatile memory and stores a segment pattern image PP1 divided into a plurality of segment areas PA1, PA2,..., PAn, as shown in FIG. 2. As described above, the segmented area pattern image PP1 stored in the pattern storage unit 130 is disposed on one side of the vehicle 200 at a height of 2.5 to 3.5 meters above the ground where the wheels 200a are in contact. The width (WP) perpendicular to the side of the vehicle 200 in one side of the image captured by the camera module 120 is at least 2.5 to 3.5 m, and the length (LP) horizontal to the side of the vehicle 200 is 5.5 to 3.5 m. It is set to 6.5m. Specifically, based on the surface of the front door 210 of the vehicle 200, the width (WP) perpendicular to the surface of the front door 210 is preferably 3 m and the length (LP) horizontal to the surface of the front door 210. is set to 6m and then divided into multiple division areas (PA1, PA2,...,PAn) in the longitudinal direction (X) or width direction (Y) of the vehicle 200.

The segment pattern image PP1 stored in the pattern storage unit 130 uses one of a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image. The vertical segment pattern image, the horizontal segment pattern image, and the semicircular segment pattern image each have a width (WP) perpendicular to the side of the vehicle 200 of 2.5 to 3.5 m and are horizontal to the side of the vehicle 200. The length (LP) is set to be included in an area of 5.5 to 6.5 m, and then the vehicle 200 is divided into a plurality of sections in the width direction (Y) or the longitudinal direction (X), respectively, to form a plurality of divided areas (PA1, PA2,...). ,PAn) is set to include,

The segment pattern image PP1 is set to the first to nth segment areas PA1, PA2,..., PAn, as shown in FIGS. 1 and 3 to 5. For example, as shown in FIGS. 3 to 5, the segment pattern image PP1 is divided into first to third segment areas PA1, PA2, and PA3, and is divided into a plurality of segment areas PA1, PA2,..., As the segment pattern image (PP1) having PAn), the vertical segment pattern image shown in FIG. 3, the horizontal segment pattern image shown in FIG. 4, and the semicircular segment pattern image shown in FIG. 5 are used, respectively. .

The vertical segmented area pattern image shown in FIG. 3 can be used to prevent accidents by detecting people in the vehicle's blind spots. The horizontal segment pattern image shown in FIG. 4 can be used to prevent safety accidents caused by approaching the vehicle 200 from behind to the front at a station. The semicircular segment pattern image shown in FIG. 5 can be used to detect a person approaching an intersection when the vehicle 200 turns right.

As shown in FIG. 2, the image overlapping unit 140 receives an image of one side of the vehicle and the segment pattern image PP1 and outputs an overlapping image in which the segment pattern image PP1 is superimposed on the image of one side of the vehicle. The image overlapping unit 140 receives the image of one side of the vehicle from the camera module 120 and the segment pattern image (PP1) output from the pattern storage unit 130, and superimposes the segment pattern image (PP1) on the image of one side of the vehicle. An overlapping image is output, and the image overlapping unit 140 processes the image using a digital image signal processing element (not shown), so description is omitted.

As shown in FIG. 2, the controller 150 receives the overlapping image and transmits it to the display 152 to display the overlapping image, and receives the location information of the vehicle 200 to calculate the moving speed of the vehicle and calculate the moving speed of the vehicle. If is a preset range, the overlapping image is analyzed to check whether there are objects in multiple segment areas (PA1, PA2,..., PAn) of the segment pattern image (PP1). If there is an object in a plurality of partitions (PA1, PA2,..., PAn), the controller 150 checks whether the object is a moving object, and if it is a moving object, turns the speaker on/off according to the partition where the moving object is located. off) The signal cycle is set differently to control and output the alarm sound output through the speaker 152.

When analyzing overlapping images, the controller 150 uses a machine learning method to identify objects in multiple partitions (PA1, PA2,...,PAn) of the segment pattern image (PP1) and the objects are moving. Check whether it is a fixed object or not. The controller 150 uses machine learning to check whether the object is a moving object or a fixed object, and if it is a moving object, it sends a speaker on-off signal according to the partition area where the moving object is located. Control the alarm sound output through the speaker by setting the cycle differently. This machine learning applies Deep Learning or Bayesian Ridge to check whether the objects located in multiple partitions (PA1, PA2,..., PAn) are moving or fixed objects, If it is a moving object, it can be judged more accurately and reliably by learning whether it is a vehicle, motorcycle, bicycle, or person.

The controller 150 also receives location information of the vehicle 200 and, according to the road conditions on which the vehicle 200 is located, includes a vertical segment pattern image among the segment pattern images PP1 stored in the pattern storage unit 130, One of the horizontal segment pattern image and the semicircular segment pattern image is output from the pattern storage unit 130 to the image overlapping unit 140, and the image overlapping unit 140 displays a vertical segment pattern image on one side of the vehicle, Controls the output of an overlapping image that overlaps either a horizontal segment pattern image or a semicircular segment pattern image.

As shown in FIG. 2, the display 151 receives and displays the overlapping image output from the image overlapping unit 140. The display 151 uses a flat panel display (not shown) and is placed inside the vehicle 200 to display an overlapping image to the driver of the vehicle 200. The overlapping image is displayed when an overlapping image in which the segment pattern image (PP1) is overlapped with the image of one side of the vehicle is received.

As shown in FIG. 2, the speaker 152 generates an alarm sound while a speaker on/off signal is received. The speaker 152 is interlocked with the generation cycle of the speaker on-off signal generated by the controller 150 to generate an alarm sound, and when a person is located close to the vehicle 200, the speaker on-off signal is generated. As the cycle speed increases, the alarm sound is also displayed at a faster cycle to warn the driver of the vehicle 200. This speaker 152 is placed inside the vehicle 200 so that the driver of the vehicle 200 can recognize the alarm sound.

The display 151 and the speaker 152 are each connected to the controller 150 of the AI camera device 100 of the present invention through wired or wireless communication, and when using the display 151 with a built-in speaker 152 When an alarm sound is generated from the speaker 152, the driver of the vehicle 200 can immediately check where the moving object is located among the plurality of partition areas (PA1, PA2,...,PAn).

In another embodiment of the AI camera device 100 for preventing traffic accidents of vehicles according to the present invention, as shown in FIG. 6, the camera module 120 uses a pair of cameras 120 and 120a.

A pair of camera modules 120 and 120a each have a horizontal viewing angle of 150 to 170 degrees, and one camera module 120 of the pair of camera modules 120 and 120a is positioned on one side of the vehicle 200. The wheel 200a is disposed above the ground at a height HT of 2.5 to 3.5 m (meters). Among the pair of camera modules 120 and 120a, the other camera module 120a is disposed on the other side of the vehicle 200 at the same height as the one camera module 120 and has a width (WP) of 2.5 to 3.5 m and a length ( LP) is a one-sided image captured by including an overlapping segment pattern image (PP2) of 17 to 19 m and a wide blind spot caused by the high overall height characteristic of the vehicle 200 in the vertical direction (Z). Outputs .

For example, one camera module 120 of a pair of camera modules 120 and 120a is disposed on the upper side of the front door 210 of the vehicle 200, as shown in FIG. 6, and the other camera module 120a ) is spaced apart from one camera module 120 disposed above the front door 210 of the vehicle 200, is disposed above the rear door 220 of the vehicle 200 at the same height, and is connected to the controller 150. . Here, one camera module 120 of the pair of camera modules 120 and 120a includes the GPS module 110, the pattern storage unit 130, the image overlapping unit 140, and the controller 150 of the case 110b. It is disposed on the AI camera device 100 shown in FIG. 1 disposed inside, and the other camera module 120a is connected to the same bracket 100a as the AI camera device 100 shown in FIG. 1 (110b). ) is placed in. That is, the other camera module 120a of the pair of camera modules 120 and 120a is connected to the controller 150 to which one camera module 120 is connected and is controlled by the controller 150 to operate on the other side of the vehicle 200. is captured and the captured image is transmitted to the overlapping image unit 140 or transmitted to the display 152 for display.

When a pair of camera modules 120 and 120a are used, the pattern storage unit 130 divides the overlapping division area pattern image PP2 in the longitudinal direction (X) or width direction (Y) of the vehicle 200 and sets it. Save it to include multiple partitions (PA1, PA2,...,PAn).

The controller 150 outputs the overlapping segment area pattern image PP2 stored in the pattern storage unit 130 from the pattern storage unit 130 to the image overlapping unit 140 and overlaps it with the image of one side of the vehicle in the image overlapping unit 140. An overlapping image of the segment pattern image PP2 is output to the display 151.

The controller 150 receives the overlapping image and transmits it to the display 152 to display the overlapping image, receives the location information of the vehicle 200, calculates the moving speed of the vehicle 200, and calculates the moving speed of the vehicle 200 so that the moving speed of the vehicle is preset. If it is a range, the overlapping image is analyzed and there is an object in multiple segment areas (PA1, PA2,...,PAn) of the overlapping segment pattern image (PP2), and if the object is a moving object, the speaker is selected according to the segment area where the moving object is located. The on-off signal cycle is set differently to control and output the alarm sound output through the speaker 152.

In this way, the controller 150 monitors the blind spot of the vehicle 200 by using one camera module 120 and the other camera module 120a among the pair of camera modules 120 and 120a to move objects. It is an object that moves and allows the driver to check whether this moving object causes a contact accident at the rear of the vehicle 200.

Another embodiment of the pattern storage unit 130 stores a segment pattern image PP1 and a plurality of inverted pattern images that invert the segment pattern image PP1, and the segment pattern images PP1 are of different colors. It includes a plurality of segment areas (PA1, PA2,…,PAn) filled with , and each of the multiple inverted pattern images contains a plurality of segment areas (PA1, PA2,…,PAn) identical to the segment pattern image (PP1). and one of the multiple partition areas (PA1, PA2,...,PAn) is set and stored transparently.

The segment pattern image PP1 is set to the first to nth segment areas (PA1, PA2,...,PAn: where n is a natural number) filled with different colors as shown in FIGS. 1 and 3 to 5. For example, as shown in FIGS. 3 to 5, the segment pattern image PP1 is divided into first to third segment areas PA1, PA2, and PA3, and the first segment area PA1 is filled with red. The second division area (PA2) of the ground is filled with yellow and the third division area (PA3) is filled with green in different colors.

The segment pattern image shown in FIG. 3 is inverted when the first segment area (PA1) is filled with red, the second segment area (PA2) is filled with yellow, and the third segment area (PA3) is filled with green. Three pattern images are provided. The three inversion pattern images have the same three segment areas as the segment pattern image shown in FIG. 3, and in one of the three inversion pattern images, the first segment area (PA1) is set from green to transparent, The remaining second divided area PA2 and third divided area PA3 are filled with the same color as shown in FIG. 3. In this way, the segment pattern image PP1 shown in FIGS. 4 and 5 is set to be the same as the reverse pattern image of the segment pattern image PP1 shown in FIG. 3.

As such, the segment pattern image (PP1) having multiple segment areas (PA1, PA2,...,PAn) filled with different colors is the vertical segment pattern image shown in FIG. 3 and the horizontal segment pattern image shown in FIG. 4. The segment pattern image and the semicircular segment pattern image shown in FIG. 5 are used.

The vertical segmented area pattern image shown in FIG. 3 can be used to prevent accidents by detecting people in the vehicle's blind spots. The horizontal segment pattern image shown in FIG. 4 can be used to prevent safety accidents caused by approaching the vehicle 200 from behind to the front at a station. The semicircular segment pattern image shown in FIG. 5 can be used to detect a person approaching an intersection when the vehicle 200 turns right.

When the pattern storage unit 130 is used in another embodiment, the image overlapping unit 140 includes an image of one side of the vehicle from the camera module 120 and a segment pattern output from the pattern storage unit 130, as shown in FIG. The image (PP1) or the inversion pattern image is received respectively, and an overlapping image in which the segment pattern image (PP1) or the inversion pattern image is superimposed on the image of one side of the vehicle is output and displayed on the display 151 under the control of the controller 150. .

The controller 150 receives the location information of the vehicle 200 output from the GPS module 110 and, if the moving speed of the vehicle 200 is within a preset range, creates a segmentation area on the image of one side of the vehicle output from the image overlay unit 140. If the pattern image (PP1) receives an overlapping image and there is an object in multiple segment areas (PA1, PA2,..., PAn) and the object is a moving object, the segment pattern image (PP1) stored in the pattern storage unit 130 ) and the divided area where the moving object is located is controlled so that the transparent inverted pattern image is replaced and output.

The controller 150 generates a segment pattern image ( PP1) and the reverse pattern image are output by increasing the replacement cycle speed so that the driver can check them through the display 151.

As shown in FIG. 7, the AI camera control method using the AI camera device 100 to prevent traffic accidents of the present invention first calculates the vehicle moving speed using the location information of the vehicle 200 (S110).

Before calculating the vehicle moving speed, the controller 150 receives the location information of the vehicle 200 and stores it in the pattern storage unit 130 according to the road conditions on which the vehicle 200 is located using the map information pre-stored in the controller 150. Among the stored segment pattern images PP1, one of a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image is output from the pattern storage unit 130. For example, the controller 150 selects a vertical segment pattern image and a horizontal segment pattern image from the pattern storage unit 130 according to road conditions such as whether the vehicle 200 is located at an intersection or a station on the road. Controls whether one of the pattern image and the semicircular division pattern image is output.

The controller 150 outputs one of the vertical segment pattern image, the horizontal segment pattern image, and the semicircular segment pattern image stored in the pattern storage unit 130 to the image overlap unit 140. ), an overlapping image that overlaps one of a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image on the image of one side of the vehicle is output.

The vehicle moving speed is calculated using the same method as described above, so description is omitted. When the moving speed of the vehicle 200 is calculated, it is checked whether the moving speed of the vehicle 200 is within the set range (S120). This means that in the case where the vehicle 200 is a route bus, the frequency with which it advances to the station is frequent, so the fatigue of the driver of the vehicle 200 from driving can be reduced by detecting it only at a certain speed or higher.

The method of entering the setting range of the moving speed of the vehicle 200 is a Wi-Fi device that includes a connector (not shown) connecting a Wi-Fi module (160) to the controller 150, and the terminal 170 is connected to the connector via a wireless communication network. Connect the module (wifi module: 160) to input the moving speed of the vehicle 200. The terminal 170 uses a laptop or tablet capable of Wi-Fi communication, connects the Wi-Fi module 160 to the connector of the controller 150 using a Wi-Fi communication network, and then connects the input device ( Set by inputting (not shown).

The specific method of connecting the controller 150 and the terminal 170 of the AI camera device 100 for preventing traffic accidents of a vehicle according to the present invention is: first, the controller 150 is equipped with a Wi-Fi communication function, and the Wi-Fi communication function is The equipped controller 150 stores a Wi-Fi SSID (Service Set Identifier) according to the Wi-Fi communication function, and when the terminal 170 is connected to the controller 150 through the Wi-Fi module 160, the terminal 170 is connected to the Wi-Fi communication function. The SSID is found and communicated with the controller 150.

In this way, the controller 150 is connected to the terminal 170 through a wireless communication network and receives setting information or road information for controlling the AI camera device 100 for preventing traffic accidents of the vehicle of the present invention and stores the pattern storage unit ( 130) or in a separate non-volatile memory (not shown), receive the GPS signal from the GPS module 110, check the location information of the vehicle 200, and then create a vertical segment pattern image and horizontal segment according to the road conditions. Select one of the area pattern image and the semicircular segment pattern image to check and monitor the object.

If the moving speed of the vehicle 200 is within the set range, check whether there is an object in the overlapping image where the segment pattern image (PP1) containing multiple segment areas (PA1, PA2,..., PAn) is overlapped on the image on one side of the vehicle. Do it (S130).

As a result of the confirmation, the controller 150 uses machine learning to determine where the object is located among the plurality of partition areas (PA1, PA2,..., PAn) and whether the object is a moving object or a fixed object. (S140). The machine learning method is Deep Learning or Bayesian Ridge, and objects are located in multiple partitions (PA1, PA2,…,PAn), and objects are separated from vehicles, motorcycles, bicycles, and people. Check whether it is a moving object or a fixed object such as a fixed signal on the road.

If the object is a moving object, the alarm generation speed is determined according to the partition area in which the object is located (S150). The controller 150 controls the alarm generation speed, that is, the generation cycle of the speaker on-off signal, according to the object being a moving object and the partition in which the object is located, and generates an alarm sound through the speaker 152 to alert the driver. .

For example, the controller 150 checks whether the moving object is located in the first partition (PA1) among the plurality of partition areas (PA1, PA2, and PA3) (S151), and determines whether the moving object is located in the first partition (PA1). When located in the area (PA1), an alarm is generated at the first speed (S152). If the moving object is not located in the first partition area (PA1), it is checked whether the moving object is located in the second partition area (PA2) among the plurality of partition areas (PA1, PA2, and PA3) (S153).

When a moving object is located in the second partition area (PA2), an alarm is generated at a second speed that is slower than the first speed (S154). If the moving object is not located in the second partition area (PA2), it is checked whether the object is located in the third partition area (PA3) among the plurality of partition areas (PA1, PA2, and PA3) (S155). When a moving object is located in the third partition area (PA3), an alarm is generated at a third speed that is slower than the second speed (S156).

When the plurality of partition areas (PA1, PA2, PA3) is used, the first partition area (PA1) is set closer to the side of the vehicle 200 than the second partition area (PA2). , the second partition area (PA2) is set closer to the side of the vehicle 200 than the third partition area (PA3). That is, the first partition area (PA1) is closer to the front door 210 of the vehicle 200, and the second partition area (PA2) is closer to the front door 210 of the vehicle 200 than the third partition area (PA3). It is set to be.

When a moving object is detected in at least two of the plurality of partitions (PA1, PA2, and PA3), the controller 150 sets the alarm generation speed based on the partition closest to the side of the vehicle 200. and print it out.

The above-described steps (S110 to S150) are controlled by the controller 150. The controller 150 controls the above-described steps (S110 to S150) according to the road conditions on which the vehicle 200 is located and whether the moving speed of the vehicle 200 is within the set range section, that is, within the set range, as shown in FIGS. 3 to 3. The object moves by selecting one of the vertical segment pattern image, horizontal segment pattern image, and semicircular segment pattern image, which are the segment pattern image PP1 stored in the pattern storage unit 130 shown in 5, respectively. Check whether it is an object.

For example, if the vehicle 200 is in a position where the vehicle 200 is traveling straight on the road and the moving speed is within the set range, the controller 150 selects a vertical segment pattern image as the segment pattern image PP1 and creates a vertical segment pattern. By checking whether there is a moving object in the image, the blind spot of the vehicle 200 is monitored so that the driver of the vehicle 200 can drive safely.

The vertical division area pattern image is divided in the width direction (Y) perpendicular to the side of the vehicle 200 to create a plurality of division areas. For example, the vertical division pattern image is divided in the width direction (Y) of the vehicle 200, which is perpendicular to the surface of the front 210, based on the front door 210 of the vehicle 200, and is divided into a plurality of division regions. This is set.

The controller 150 selects this vertical segment pattern image to allow the driver to recognize whether a moving object is approaching the vehicle 200 when the vehicle 200 is straight on the road. For example, the controller 150 checks whether a moving object approaches from a partition farther from the side of the vehicle 200 to a partition closer to the side of the vehicle 200 among a plurality of partitions included in the vertical partition pattern image, and determines whether the object is It is possible to check whether it is a moving object and how close the moving object is, making it possible to easily check whether there is a moving object in the blind spot of the vehicle 200 and prevent vehicle accidents.

If the vehicle 200 is approaching the intersection and the moving speed is within the set range, the controller 150 selects the semicircular segment pattern image as the segment pattern image PP1 when turning right at the intersection and displays the area within the semicircular segment pattern image. It checks whether there is a moving object and monitors whether there is a moving object crossing the crosswalk installed at the intersection when the vehicle 200 turns right, so that the driver of the vehicle 200 can drive safely.

The semicircular division area pattern image is divided into a semicircle in the width direction (Y) perpendicular to the side of the vehicle 200, thereby establishing a plurality of division areas. For example, the semicircular division pattern image is divided into a semicircle in the width direction (Y) perpendicular to the surface of the vehicle 200 based on the side of the vehicle 200, that is, the front door 210, and is divided into a plurality of divisions. Since the area is set, when the vehicle 200 turns right at a road intersection, objects approaching the front door 210 of the vehicle 200, that is, people, can be identified, allowing the vehicle 200 to turn right and pass safely.

As described above, the controller 150 selects a semicircular segment pattern image so that when the vehicle 200 turns right, the object moving toward the vehicle 200 is selected from a segment far from the vehicle 200 among multiple segment areas to a segment close to the segment area. By detecting whether the vehicle 200 is approaching, it is possible to monitor whether there is a moving object crossing the crosswalk installed at the intersection when the vehicle 200 turns right. The semicircular segment pattern image identifies whether the moving object is a motorcycle, bicycle, or person.

If the location of the vehicle 200 enters or departs from a road stop and the moving speed is within the set range, the controller 150 selects a horizontal segment pattern image as the segment pattern image (PP1) and creates a horizontal segment pattern image. Check whether there is a moving object inside the vehicle 200 and monitor whether there is an object moving from the back of the vehicle 200 to the front door 210 when the vehicle 200 enters or departs the station so that the driver of the vehicle 200 can drive safely. do.

The horizontal division pattern image is divided in the longitudinal direction (X), which is horizontal to the side of the vehicle 200, to create a plurality of division regions. In other words, the horizontal division pattern image is divided in the longitudinal direction ( can be confirmed in a wide range.

When the location of the vehicle 200 enters or departs from a road stop, the controller 150 selects this horizontal segment pattern image and determines whether there is a moving object in a plurality of segment areas included in the horizontal segment pattern image. It is possible to prevent vehicle accidents that may occur when the vehicle 200 enters or exits the station by detecting whether an object is approaching from a division farther from the vehicle 200 to a division closer to the vehicle 200 among the plurality of divisions. The horizontal segment pattern image can detect many moving objects gathered in the longitudinal direction (X) of the vehicle 200 at a station, that is, people entering or leaving the vehicle 200.

The controller 150 outputs the vertical partition pattern image, horizontal partition pattern image, and semicircular partition pattern image stored in the above-described pattern storage unit 130 as an overlapping partition pattern image (PP2), respectively, as shown in FIG. 6. As shown in the vertical division pattern image, they are set to overlap in the longitudinal direction (X) of the vehicle 200. For example, when a vertical segment pattern image is applied, the overlapping segment pattern image PP2 is divided in the width direction (Y) of the vehicle 200, which is perpendicular to the side of the vehicle 200, as shown in FIG. 6. It is set to be divided in the longitudinal direction (X) of the vehicle 200 when applying a horizontal segment pattern image, and is set to be symmetrical in the longitudinal direction (X) of the vehicle 200 when applying a semicircular segment pattern image. It is used when applying a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image, respectively, by extending them from one side of the vehicle 200 to the other side.

As described above, the AI camera device and its control method for preventing traffic accidents of vehicles of the present invention divide one side or the other side of the vehicle into a plurality of partition areas and photograph them with a camera, and a plurality of images captured according to the moving speed of the vehicle are used. After checking whether there is a moving object in the partition area, if there is a moving object, the warning generation can be controlled depending on where it is located among the plurality of partition areas, thereby preventing vehicle traffic accidents.

The AI camera device and its control method for preventing traffic accidents of vehicles of the present invention are applied to the camera manufacturing industry or industrial fields that manufacture safety accident prevention devices.

100: AI camera device 110: GPS module
120: Camera module 130: Pattern storage unit
140: Image overlapping part 150: Controller
151: Display 152: Speaker
200: vehicle 210: front door
220; back door

Claims (8)

A GPS module that receives GPS (Global Positioning System) signals and outputs location information of the vehicle,
A camera module that photographs one side of the vehicle and outputs an image of one side of the vehicle,
a pattern storage unit that stores a segment pattern image divided into a plurality of segment areas;
an image overlay unit that receives the image of one side of the vehicle and the segment pattern image and outputs a superimposed image in which the segment pattern image is superimposed on the image of one side of the vehicle;
The overlapping image is received and transmitted to the display to display the overlapping image, and the location information of the vehicle is received to calculate the moving speed of the vehicle. If the vehicle moving speed is within a preset range, the overlapping image is analyzed to determine the segment pattern image. If there are objects in multiple partitions and the object is a moving object, the cycle of the speaker on-off signal is set differently depending on the partition where the moving object is located, and the alarm sound output through the speaker is controlled and output. It includes a controller that does,
The camera module is placed on one side of the vehicle at a height of 2.5 to 3.5 m above the ground where the wheels are in contact, captures one side to include the segment pattern image, and outputs a captured image of one side,
The segment pattern image stored in the pattern storage unit uses one of a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image, and the vertical segment pattern image and the horizontal segment pattern image are used. and the semicircular segmented area pattern image is set to include an area of 2.5 to 3.5 m in width perpendicular to the side of the vehicle and 5.5 to 6.5 m in length horizontal to the side of the vehicle, and then divided in the width or length direction, respectively. It is divided and set to include multiple partitions,
The controller receives the location information of the vehicle and selects a vertical segment pattern image, a horizontal segment pattern image, and a semicircular segment pattern image among the segment pattern images stored in the pattern storage unit according to the road condition on which the vehicle is located. Output one of them from the pattern storage unit to the image overlapping unit to output an overlapping image in which one of the vertical segment pattern image, horizontal segment pattern image, and semicircular segment pattern image is superimposed on the image of one side of the vehicle. AI camera device to prevent traffic accidents in vehicles that are driving. According to paragraph 1,
The camera module is used with a horizontal viewing angle of 150 to 170 degrees,
An AI camera device for preventing traffic accidents of vehicles with a height of 2.0 to 4.0 m (meters) and a length of 10 to 13 m from the ground where the wheels come into contact. According to paragraph 1,
The controller uses a machine learning method when analyzing overlapping images to determine whether there is an object in multiple partitions of the segment pattern image and whether the object is a moving object or a fixed object. If it is a moving object, the controller determines whether the object is moving or not. Controls the alarm sound output through the speaker by setting the cycle of the speaker on-off signal differently depending on the partition area where the object is located.
The machine learning is an AI camera device for preventing traffic accidents in vehicles where Deep Learning or Bayesian Ridge is applied. delete According to paragraph 1,
The vertical division area pattern image is divided in the width direction perpendicular to the side of the vehicle to create a plurality of division areas,
The horizontal division pattern image is divided in the longitudinal direction parallel to the side of the vehicle to create a plurality of division regions,
The semicircular segmented area pattern image is divided into a semicircle in the width direction perpendicular to the side of the vehicle, thereby establishing a plurality of segmented areas. An AI camera device for preventing traffic accidents in vehicles. According to paragraph 1,
The camera module uses a pair of camera modules,
Of the pair of camera modules, one camera module is placed on one side of the vehicle at a height of 2.5 to 3.5 meters (meters) above the ground where the wheels are in contact, and the other camera module is placed on the other side of the vehicle. A single-sided image is output by shooting an image containing an overlapping segment pattern that is arranged at the same height and has a width of 2.5 to 3.5 m and a length of 17 to 19 m,
The pattern storage unit stores the overlapping division area pattern image to include a plurality of division areas set by dividing the vehicle in the longitudinal or width direction,
The controller outputs the overlapping segment pattern image stored in the pattern storage unit from the pattern storage unit to the image overlapping unit, receives an overlapping image in which the overlapping segment pattern image is superimposed on the image of one side of the vehicle from the image overlapping unit, and outputs it to the display. , receives the location information of the vehicle, calculates the moving speed of the vehicle, analyzes the overlapping image if the moving speed of the vehicle is within a preset range, and moves if there is an object in multiple divided areas of the overlapping segment pattern image and the object is a moving object. An AI camera device to prevent traffic accidents in vehicles that controls and outputs alarm sounds through speakers by setting the cycle of the speaker on-off signal differently depending on the partition area where the object is located. delete delete

Images