[go: up one dir, main page]

WO2017078213A1 - Procédé pour détecter un objet en déplacement dans une image photographiée, et système de prévention d'accident d'embarquement et de débarquement l'utilisant - Google Patents

Procédé pour détecter un objet en déplacement dans une image photographiée, et système de prévention d'accident d'embarquement et de débarquement l'utilisant Download PDF

Info

Publication number
WO2017078213A1
WO2017078213A1 PCT/KR2015/013119 KR2015013119W WO2017078213A1 WO 2017078213 A1 WO2017078213 A1 WO 2017078213A1 KR 2015013119 W KR2015013119 W KR 2015013119W WO 2017078213 A1 WO2017078213 A1 WO 2017078213A1
Authority
WO
WIPO (PCT)
Prior art keywords
light flow
vehicle
region
frequency
area
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/KR2015/013119
Other languages
English (en)
Korean (ko)
Inventor
양승한
박성령
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Industry Academic Cooperation Foundation of KNU
Original Assignee
Industry Academic Cooperation Foundation of KNU
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Industry Academic Cooperation Foundation of KNU filed Critical Industry Academic Cooperation Foundation of KNU
Publication of WO2017078213A1 publication Critical patent/WO2017078213A1/fr
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Images

Classifications

    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60N—SEATS SPECIALLY ADAPTED FOR VEHICLES; VEHICLE PASSENGER ACCOMMODATION NOT OTHERWISE PROVIDED FOR
    • B60N5/00—Arrangements or devices on vehicles for entrance or exit control of passengers, e.g. turnstiles
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60Q—ARRANGEMENT OF SIGNALLING OR LIGHTING DEVICES, THE MOUNTING OR SUPPORTING THEREOF OR CIRCUITS THEREFOR, FOR VEHICLES IN GENERAL
    • B60Q5/00—Arrangement or adaptation of acoustic signal devices
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
    • B—PERFORMING OPERATIONS; TRANSPORTING
    • B60—VEHICLES IN GENERAL
    • B60R—VEHICLES, VEHICLE FITTINGS, OR VEHICLE PARTS, NOT OTHERWISE PROVIDED FOR
    • B60R21/00—Arrangements or fittings on vehicles for protecting or preventing injuries to occupants or pedestrians in case of accidents or other traffic risks
    • B60R21/01—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents
    • B60R21/013—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting collisions, impending collisions or roll-over
    • B60R21/0134—Electrical circuits for triggering passive safety arrangements, e.g. airbags, safety belt tighteners, in case of vehicle accidents or impending vehicle accidents including means for detecting collisions, impending collisions or roll-over responsive to imminent contact with an obstacle, e.g. using radar systems
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B21/00—Alarms responsive to a single specified undesired or abnormal condition and not otherwise provided for
    • G08B21/18—Status alarms

Definitions

  • the present invention is to prevent accidents when getting on and off the vehicle, in particular by comparing the surrounding images of the vehicle to detect the moving object based on the light flow derived by the moving object, to more accurately guide the dangerous state around the vehicle
  • the present invention relates to a method for detecting a moving object in a captured image and a system for preventing getting on and off a vehicle using the same.
  • the driver visually checks the surroundings of the vehicle through an external side mirror and an interior room mirror provided at the left / right side of the vehicle while driving or parking to prepare for an accident.
  • cameras are installed in vehicles to facilitate the driver's monitoring of the surroundings of the vehicle from the inside of the vehicle, as well as dangerous situations such as people approaching the vehicle.
  • a safety device in a vehicle that performs a function such as generating an alarm sound inside the vehicle.
  • the safety device compares the previous photographed image and the current photographed image provided by the camera installed in the vehicle to determine the state of the surroundings of the vehicle, particularly moving objects such as various transport means such as a moving vehicle or a motorcycle. .
  • the safety device determines the moving object using only the change of the captured image, there is a limit that it is possible to guide the dangerous state of the surrounding situation only when the vehicle is stopped.
  • the present invention was created in view of the above circumstances, and compares the surrounding images of the vehicle to verify the same object based on the light flow of the changed object, and the light flow map corresponding to the movement change between the corresponding object feature points.
  • the technical purpose is to provide a vehicle unloading accident prevention system used.
  • the present invention is to determine the current state of the vehicle based on the light flow distribution region of the light flow map, and to determine the moving object corresponding to the state of the vehicle based on the light flow distribution region for the preset region of interest. Accordingly, another technical object is to provide a method for detecting a moving object in a captured image and a vehicle getting on and off accident prevention system using the same, which can recognize a dangerous situation regardless of the state of the vehicle.
  • a photographing apparatus installed in the vehicle for photographing the outer periphery of the vehicle, a notification device for outputting accident prevention notification information, and an object in the image frame provided from the photographing apparatus Detects the feature points for, generates a light flow vector corresponding to the movement of the feature points in the current image frame and the previous image frame, discretizes the light flow vector with respect to the size and angle,
  • a vehicle getting on and off accident prevention system comprising a control device for controlling to transmit the alarm information corresponding to determine that the moving object exists through the notification device.
  • the control apparatus may further include a feature point detection module for detecting a corner feature point of an object in a current image frame, and a feature point of a previous or subsequent image frame corresponding to a corner feature point in the current image frame detected by the feature point detection module.
  • a feature point detection module for detecting a corner feature point of an object in a current image frame, and a feature point of a previous or subsequent image frame corresponding to a corner feature point in the current image frame detected by the feature point detection module.
  • the light flow vector including the size corresponding to the moving change distance of the object and the angle corresponding to the moving direction.
  • a light flow vector generation module for generating a light flow map and a light flow map having a light flow size axis and an angular axis based on the light flow vector provided from the light flow vector generation module
  • An optical flow map generation module for calculating a flow frequency and an optical flow map generated by the optical flow generation module
  • the light flow vector generation module is provided with a vehicle getting on and off accident prevention system, characterized in that configured to generate a light flow vector from the image of the upper line of view and the light flow coordinate change processing.
  • the light flow vector generation module is provided with a vehicle unloading accident prevention system, characterized in that configured to set only the light flow vector of the light flow vector of the size of the light flow vector less than a predetermined effective reference pixel as the effective light flow vector.
  • the motion analysis module determines the stop, forward and backward state of the vehicle on the basis of the light frequency region of the maximum frequency in the light flow map, but the minimum size of the light flow in the light flow map consisting of the size axis and the angular axis Set the area less than the pixel as the stationary state area, set the area where the change of angle as the minimum reference angle range is small, and set the area where the change of angle as the maximum reference angle range as the reverse state area.
  • a vehicle unloading accident prevention system characterized in that configured to determine the movement state of the vehicle.
  • the light flow map generation module generates a first light flow map including light flow frequency for each region based on the light flow vector for the entire image frame, and the light flow vector for the region of interest preset in the image frame. Generate a second light flow map including light frequency by region based on the second light flow frequency; and the motion analysis module analyzes the first light flow map to determine a stop, forward, and reverse states of the vehicle, and Provided is a vehicle unloading accident prevention system, characterized in that the light flow map is analyzed to determine the existence of a moving object.
  • the motion analysis module may be configured to determine that the motion object exists when the highest frequency region in the first light flow map and the highest frequency region in the second light flow map are different from each other when the motion object is not detected.
  • a vehicle unloading accident prevention system is provided.
  • the motion analysis module may cause an object to be jammed if the light flow continues to be detected at a predetermined frequency or more in a region other than the region where the angle change is the maximum reference angle range and the region where the angle change is the minimum reference angle range in the light flow map.
  • the photographing device is installed in the vehicle for photographing the outer periphery of the vehicle, vehicle state information providing means for providing the current state of the vehicle, accident prevention notification A notification device for outputting information, a feature point for the object in the region of interest of the captured image provided from the imaging apparatus, and a light flow vector corresponding to the change in the feature point movement in the region of interest of the current image frame and the previous image frame; And generating a discretized light flow vector with respect to size and angle, and determining that a moving object exists when the maximum frequency in an area other than the vehicle state area corresponding to the vehicle state provided from the vehicle state information providing means is equal to or greater than a reference frequency. Controlling to transmit alarm information corresponding thereto through the notification device
  • the getting on and off the vehicle accident prevention system characterized in that comprises a device is provided.
  • the control apparatus may further include a feature point detection module for detecting a corner feature point of an object in a current image frame, and a feature point of a previous or subsequent image frame corresponding to a corner feature point in the current image frame detected by the feature point detection module.
  • a feature point detection module for detecting a corner feature point of an object in a current image frame, and a feature point of a previous or subsequent image frame corresponding to a corner feature point in the current image frame detected by the feature point detection module.
  • the light flow vector including the size corresponding to the moving change distance of the object and the angle corresponding to the moving direction.
  • a light flow vector generation module for generating a light flow map and a light flow map having a light flow size axis and an angular axis based on the light flow vector provided from the light flow vector generation module
  • An optical flow map generation module for calculating a flow frequency and an optical flow map generated by the optical flow generation module
  • a vehicle unloading accident prevention system comprising a motion analysis module for determining that a moving object exists when the maximum light flow frequency in a region other than the region corresponding to the current state of the vehicle is higher than the reference frequency.
  • a fourth step a moving object is detected in the captured image characterized in that the configuration including the fifth step of detecting the moving object is an optical flow frequency in the fourth step on
  • the method of detecting a motion object in the photographed image may include performing an upward gaze coordinate transformation process on the image including the light flow information and generating an upward gaze light flow vector from the upward gaze image. Is provided.
  • the motion object in the photographed image may be configured to set only the light flow vector having the size of the light flow vector less than the preset effective reference pixel as the effective light flow vector and then apply to the detection of the motion object.
  • a detection method is provided.
  • the third step generates a first light flow map of the light flow vector of the entire image frame, and generates a second light flow map based on the light flow vector of the region of interest preset in the image frame.
  • the fourth step calculates the light flow frequency for each region for each of the first and second light flow maps, and the fifth step is the maximum light flow frequency calculated by the first light flow map.
  • the fifth step may include the vehicle calculated by the first light flow map when an area in which the light flow frequency calculated by the second light flow map is equal to or greater than a reference frequency does not exist in an area other than the vehicle state area.
  • a moving object detection method in a captured image is provided that determines whether a moving object exists in a moving object determination region corresponding to a state.
  • the region in which the light flow is less than the reference pixel is set as the stationary region in the first light flow map including the size axis and the angle axis, and the area having the smallest change in angle is moved as the minimum reference angle range.
  • a moving object detection method in a captured image characterized in that it is set to a state area, and the area having a large change in angle as a maximum reference angle range is set as a reverse state area to determine a current state of the vehicle.
  • the motion object when the motion object is not detected, it is determined that the motion object exists when the highest frequency region in the first light flow map and the highest frequency region in the second light flow map are different from each other.
  • a moving object detection method in a captured image is provided.
  • the jammed object is generated when the light flow continues to be detected at a predetermined frequency or more in a region other than the region where the angle change is the maximum reference angle range and the region where the angle change is the minimum reference angle range.
  • the feature point for the object in the 51st step of receiving the current state of the vehicle from the vehicle, and the region of interest of the captured image provided from the imaging device installed in the vehicle In step 52, the ROI feature points of the previous image frame are estimated based on the ROI feature points of the current image frame detected in step 52, and the moving direction is based on the positional relationship between the ROI feature points of the two image frames.
  • the light flow frequency of the predetermined area is calculated from the light flow map generated in step 54 and the light is emitted from an area other than the area corresponding to the vehicle state provided in step 51 from step 55.
  • a motion object detection method in a captured image is provided, comprising the step 56, determining that there is a motion object when a region having a flow frequency greater than or equal to the reference frequency exists.
  • the stop state is set to an area where the magnitude of the light flow is less than the reference pixel in the light flow map, and the forward state is set to an area having a small change in angle as a preset minimum reference angle range.
  • the present invention by comparing the surrounding image of the vehicle and verifying the same object based on the light flow of the changed object by detecting the moving object, it is more accurate to recognize the moving object around the vehicle to accurately guide the danger state accordingly I can give it.
  • the current vehicle state is determined based on the light flow distribution region of the light flow map, and the moving object corresponding to the state of the vehicle is determined based on the light flow distribution region for the preset ROI.
  • FIG. 1 is a view showing a schematic configuration of a vehicle unloading accident prevention system according to the present invention.
  • FIG. 2 is a diagram illustrating the internal structure of the control device 400 shown in FIG. 1 by function.
  • FIG. 3 is a view for explaining a feature point detection operation in the feature point detection module 410 shown in FIG.
  • FIG. 4 is a view for explaining an operation of generating a light flow vector of the light flow vector generating module 420 shown in FIG.
  • FIG. 5 is a diagram illustrating a light flow map generated by the light flow map generation module 430 shown in FIG. 2.
  • 6 to 9 are views for explaining a motion state determination operation in the motion object analysis module 440 shown in FIG.
  • FIGS. 10 and 11 are flowcharts illustrating a method of detecting a motion object in a captured image according to the present invention.
  • FIG. 1 is a view showing a schematic configuration of a vehicle unloading accident prevention system according to the present invention.
  • the vehicle getting on and off accident prevention system includes a photographing apparatus 100, an alarm device 200, a data memory 300, and a control device 400.
  • the photographing apparatus 100, the alarm apparatus 200, the data memory 300, and the control apparatus 400 are installed at a predetermined position of the vehicle.
  • the photographing apparatus 100 is installed in a vehicle and photographs the outer periphery of the vehicle, and may be configured as a camera.
  • the photographing apparatus 100 is installed at an appropriate position outside the vehicle, for example, a side mirror or an upper side of the door so as to photograph the side of the door of the vehicle.
  • the photographing region of the photographing apparatus 100 may be set to include a region of interest (ROI), that is, a danger region.
  • ROI region of interest
  • the dangerous area may be set to an appropriate area in consideration of the type of vehicle or the getting on and off environment, and may be set to a plurality of different areas according to the installation position of the photographing apparatus 100.
  • the danger zone may be set to an area including a door and a front wheel and a rear wheel of the vehicle.
  • the alarm device 200 may be configured as, for example, a buzzer or a speaker as a device for outputting an alarm sound or a guide sound based on the information provided from the control device 400.
  • the alarm device 200 may be installed inside or outside the vehicle. In this case, the alarm device 200 is installed in the interior of the vehicle to recognize the accident prevention notification state to the driver and passengers located in the room, and the outside of the vehicle is installed to the passengers who get on and off the vehicle through the door of the accident prevention This is to recognize the notification status.
  • the data memory 300 stores pixel position information of the ROI in the captured image and various reference information for analyzing the motion object.
  • the reference information includes effective reference size information for setting an effective light flow vector applied to motion analysis, a reference frequency for determining a moving object, and a minimum reference angle range / maximum reference angle range / minimum reference for determining a motion state region. Pixel range is included.
  • the reference angle range is a range for determining the forward and backward movement of the vehicle
  • the reference pixel range is a range for determining the stop of the vehicle.
  • the control device 400 determines a moving object around the vehicle door regardless of the movement of the vehicle based on the captured image provided from the photographing apparatus 100, and provides alarm information corresponding to the moving object based on the determination result of the moving object. It is configured to output through the alarm device 200.
  • FIG. 2 is a diagram illustrating the internal structure of the control device 400 shown in FIG. 1 by function.
  • the control device 400 includes a feature point detection module 410, an optical flow vector generation module 420, an optical flow map generation module 430, and a motion object analysis module 440. It is composed.
  • the feature point detection module 410 detects a feature point for the current image frame provided from the photographing means 100.
  • the feature point detection module 410 may detect corner feature points for all objects O in the corresponding image frame, as shown in FIG. 3A, using a known Harris Corner detector. . In this case, the feature point detection may be performed after the image is reduced to improve the feature point detection speed.
  • 3B is a diagram illustrating a photographed image in which feature points are displayed, wherein a red circle represents a feature point.
  • the light flow vector generation module 420 analyzes the positional relationship of the corner feature points of the object detected by the feature point detection module 410 to obtain light flow information corresponding to the moving direction, and generates a light flow vector based on this. do.
  • the light flow vector generation module 420 estimates a point corresponding to a feature point of the current image frame with respect to a previous or subsequent image frame by using the Pyramidal Lucas Kanade method.
  • the light flow vector generation module 420 is an object in two frames, for example, the object O 2 in the current frame and the object O 1 in the previous frame, as shown in FIG.
  • the position change of each feature point corresponding to each other is detected, and an optical flow having a movement direction corresponding to the position change is obtained.
  • the light flow for the stationary object becomes a "0" vector.
  • the light flow vector generation module 420 converts the obtained light flow into the light flow in the upward gaze image in order to remove the effects of radiation and perspective distortion.
  • a fisheye lens camera generally used is output as a photographed image including perspective distortion according to a camera posture and radiation distortion according to lens characteristics as shown in 1 in FIG. Accordingly, the light flow vector generation module 420 may generate the light flow in a state in which the photographed image including the distortion component in the form of 1 is converted into an upward gaze image in the form of 2 using a preset correction coefficient. . In this case, it is also possible to convert only the light flow information into an upward gaze image.
  • the light flow vector generation module 420 transforms coordinates of the start and end points of the light flow by using an upward gaze transformation algorithm by using a preset correction parameter, thereby generating a new image in an upward gaze image such as a bird's eye view.
  • the light flow vector OV may be generated as shown in FIG. 4B.
  • the light flow vector generation module 420 may emit in the process of calculating the light flow information. Therefore, the light flow vector generation module 420 is preferably limited to the range in which the light flow information may be physically moved. Do.
  • the light flow vector generation module 420 is effective only for the light flow vector having the magnitude of the light flow vector generated for all the feature points including the change occurrence object, that is, the pixel difference is less than or equal to the preset effective reference size, "30 pixels". Can be set as light flow vector.
  • the light flow map generation module 430 generates a light flow map to be used to determine the state of the vehicle and the movement of the object.
  • the first light flow map for determining the state of the vehicle is generated based on the light flow vector in the entire image frame
  • the second light flow map for determining whether the object is moved is determined in the preset ROI. Generated based on the light flow vector.
  • the light flow map generation module 430 discretizes the effective light flow vector in the light flow vector generation module 420 so as to correspond to the size and angle, as shown in FIG. 5A or 5B.
  • the first and second light flow maps are generated through the light flow information calculation process. That is, the light flow map generation module 430, for example, for the three light flows of (1,0) (1,2) (1,3), as shown in FIG. A table of sizes and angles for each coordinate is generated, and a light flow distribution chart in a graph form for sizes and angles is generated as shown in FIG. 5A and 5B show data discretized with a size of "0.5" pixels and an angle of "1 °".
  • the motion object analysis module 440 determines a motion state of the vehicle on the basis of the light flow frequency for each motion state region in the first light flow map generated by the light flow map generation module 430. The existence of the moving object is determined based on the second light flow map.
  • the motion analysis module 440 first analyzes the current light state of the vehicle by analyzing the first light flow map. That is, the motion analysis module 440 calculates the frequency of the light flow for each state region as shown in FIG. 6 and stops, forwards, and reverses according to the distribution region having the maximum frequency. Analyze with (BACKWARD). In the stopped state, the light flow becomes a "0" vector because most feature point positions remain unchanged when the vehicle is stopped. In other words, the stationary region may be set to a relatively small region, for example, an region in which the size of the light flow vector is less than "1" pixel.
  • the advanced state region may be set as a region having a relatively small change in angle, that is, a region having a predetermined minimum reference angle range, for example, an area having an angle range of "-10 to +10".
  • the backward state region may be set as a region having a relatively large change in angle, that is, a region having a maximum reference angle range, for example, a region having an angle range of "-170 to -180" and "+170 to +180". That is, the motion analysis module 440 determines that the current vehicle is in the stopped state when the frequency of light flow in the stationary state region is the highest in the current photographed image.
  • the motion analysis module 440 determines that a motion object exists when there is an area in the second light flow map other than the current state of the vehicle in which the local light flow maximum frequency is greater than or equal to a preset reference frequency. For example, when the vehicle is in a stopped state, it is determined that a moving object exists when a maximum frequency exists in an area other than the stopped state area in the second light flow map.
  • the motion analysis module 440 is configured to provide the accident prevention alarm information corresponding to the alarm device 200 based on the detection result of the above-described moving object.
  • the motion analysis module 440 calculates a first frequency Total corresponding to the vehicle state in the first light flow map in FIG. 7, and a second frequency ROI corresponding to the motion object in the second light flow map. ), It is determined that a moving object exists when there is an area having the second highest frequency in an area other than the area having the highest first frequency. That is, in Fig. 6B, the vehicle state area is "X”, and the moving object area is "Y", which is the second frequency occurrence area having the highest frequency in areas other than the vehicle state area.
  • the motion analysis module 440 may be configured to determine that an object is jammed in the vehicle. In this case, when the object is caught in the vehicle, the object moves with it. Accordingly, the motion analysis module 440 considers this and the light flow in the vertical direction as shown in FIG. If it appears continuously for more than a certain time, it is determined that the jammed object has occurred.
  • the state of the vehicle is determined by using the light flow map in the entire image frame, but the state of the vehicle may be provided by the apparatus of the vehicle itself. It is possible.
  • the current state of the vehicle may be provided through vehicle state information providing means (not shown) provided in the vehicle. Therefore, according to the present invention, as described above, only the region of interest preset in the image frame may be generated to determine the motion object by generating the light flow map. In this case, a process of generating and analyzing a light flow map for determining a vehicle state may be omitted, and thus the processing speed may be improved when using the same system.
  • a moving object exists by first determining a moving object in the ROI around the vehicle through the second light flow map, an alarm for the moving object is not determined without determining the state of the vehicle.
  • the processing speed can be further improved when using the same system.
  • the vehicle state is determined through the first light flow map, and then an alarm is detected for the moving object by checking whether a moving object exists in an area corresponding to the vehicle state. It is configured to provide.
  • the current state information of the vehicle may be provided through vehicle state information providing means (not shown) provided in the vehicle, and if a moving object does not exist with respect to the current state of the vehicle, it determines the jammed object and provides an alarm. Can be done.
  • the photographing apparatus 100 provides a photographed image to the control apparatus 400 at a predetermined cycle.
  • the photographing apparatus 100 may be installed above the door of the vehicle, but the photographing apparatus may be installed to include a dangerous area around the vehicle.
  • the control device 400 detects a feature point of the object using a Harris corner detector in the current image frame (ST10).
  • the control device 400 estimates a feature point corresponding to the corner feature point of the current image frame from the previous image frame, and obtains light flow information corresponding to the positional change between the corner feature points in the two image frames (ST20). That is, as shown in (A) of FIG. 4, the control device 400 has a light flow corresponding to a position change between a corner feature point in the current image frame and a corner feature point in the previous image frame corresponding to the object. Generate information.
  • the control device 400 converts an image including light flow information into an upward line of sight image and generates a light flow vector corresponding to the light flow information in the upward line of sight image (ST30 and ST40). That is, the control device 400 generates a light flow vector including a size corresponding to the moving distance and angle information corresponding to the moving direction in the light flow information of the object. At this time, the control device 400 may set only the light flow vector of which the size is less than a predetermined size reference pixel, for example, “30” pixels, among the light flow vectors as the effective light flow vector. That is, the light flow vector larger than the reference pixel size is cleared in the corresponding upward line image.
  • the upward gaze light flow vector may be generated by coordinate transformation so that only the light flow vector information corresponds to the upward gaze.
  • the controller 400 generates a light flow map for the size and angle of each light flow vector based on the effective light flow vector (ST50). That is, the controller 400 generates a light flow table as shown in FIG. 5A, and generates a light flow distribution chart in a graph form as shown in FIG. 5B so as to correspond thereto. do. In this case, the control device 400 generates a first light flow map of light flow frequency for each region based on the light flow vector for the entire image frame, and also the light flow vector for the region of interest preset in the image frame. Generate a second light flow map consisting of the light flow frequency for each region based on the.
  • control device 400 analyzes the first light flow map to calculate the light flow frequency for each preset vehicle state region, and determines the current state of the vehicle based on this (ST60). That is, the control device 400 determines the stationary state, the forward state and the reverse state of the vehicle based on the light flow frequency in the entire image as shown in FIG. 6.
  • control device 400 determines the presence or absence of a moving object by checking the light flow frequency in the second light flow map for the preset ROI (ST70). That is, the control device 400 determines that the moving object exists when the maximum frequency is greater than or equal to the reference frequency in the remaining regions other than the current vehicle state region using only the light flow vector existing in the preset ROI.
  • control device 400 additionally determines whether there is a moving object moving in the same manner as the vehicle, or determines whether there is a jammed object in the vehicle.
  • the moving object is determined using the second light flow map.
  • the second light flow map is applied to the second light flow map. If there is no area in the area other than the vehicle state area in which the light flow frequency calculated by the first frequency is greater than or equal to the vehicle state area, there is a motion object for the motion object determination area corresponding to the vehicle state calculated by the first light flow map. It is also possible to implement to judge.
  • control device 400 uses the second light flow map to use the second light flow map other than the vehicle state determination region shown in FIG. 9. It is determined whether an area motion object exists at step ST110.
  • step ST110 If it is determined in step ST110 that the moving object exists, the control device 400 outputs accident prevention alarm information (ST120).
  • step ST110 determines whether the moving object does not exist. If it is determined in step ST110 that the moving object does not exist, the control device 400 analyzes the current state of the vehicle using the first light flow map (ST130).
  • control device 400 re-confirms whether the object is a moving object using the first and second light flow maps corresponding to the state of the vehicle analyzed in step ST130.
  • step ST130 when it is determined in step ST130 that the current vehicle is in a stopped state, the control device 400 has an accident when a moving object exists in an area except for the vehicle stopping area ((a) of FIG. 6) in the second light flow map. Output the preventive alarm information (ST140).
  • control device 400 determines that the current vehicle is in the advanced state at step ST130, when the moving object exists in an area other than the vehicle forward area (b) of the second light flow map, The accident prevention alarm information is output (ST150).
  • control device 400 when it is determined that the current vehicle is in the reverse state in step ST130, the control device 400 has an accident when there is a moving object in an area other than the vehicle reverse area ((c) of FIG. 6) in the second light flow map. Output the preventive alarm information (ST160).
  • the control device 400 when the moving object is not present in the area other than the vehicle forward area or the vehicle reverse area in the step ST150 and ST160, if the jammed object is confirmed by checking whether the vehicle jammed object exists, Output the preventive alarm information (ST170). At this time, the control device 400, as shown in FIG. If the error persists, it is determined that a jammed object has occurred.
  • the embodiment it is detected whether a moving object exists through comparison of the captured image, but the surrounding of the vehicle regardless of the state of the vehicle by using the light flow vector obtained by the change of movement with respect to the moving object in the captured image It is possible to more accurately detect the motion object detection of.

Landscapes

  • Engineering & Computer Science (AREA)
  • Mechanical Engineering (AREA)
  • Physics & Mathematics (AREA)
  • Transportation (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Automation & Control Theory (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Emergency Management (AREA)
  • Acoustics & Sound (AREA)
  • Business, Economics & Management (AREA)
  • Image Analysis (AREA)
  • Traffic Control Systems (AREA)
  • Human Computer Interaction (AREA)
  • Electromagnetism (AREA)
  • Mathematical Physics (AREA)
  • Multimedia (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)

Abstract

La présente invention concerne un procédé pour détecter un objet en déplacement dans une image photographiée, lequel procédé est apte à indiquer une situation dangereuse entourant un véhicule de façon plus précise, et un système de prévention d'accident d'embarquement et de débarquement l'utilisant. Le système de prévention d'accident d'embarquement et de débarquement comprend : un dispositif de photographie, installé sur un véhicule, pour l'environnement externe du véhicule ; un dispositif de notification pour émettre une information d'alarme de prévention d'accident ; et une unité de commande configurée pour détecter un point caractéristique d'un objet dans une trame d'image délivrée à partir du dispositif de photographie, pour générer un vecteur de flux lumineux correspondant à un changement dans le mouvement du point caractéristique dans la trame d'image actuelle et une trame d'image précédente, discrétiser le vecteur de flux lumineux vis-à-vis de la grandeur et de l'angle, déterminer qu'un objet en déplacement est présent quand la fréquence maximale dans une région autre qu'une région de détermination d'état de véhicule prédéterminée est supérieure ou égale à une fréquence de référence, et transmettre une information d'alarme correspondant à cette dernière par l'intermédiaire du dispositif de notification.
PCT/KR2015/013119 2015-11-05 2015-12-03 Procédé pour détecter un objet en déplacement dans une image photographiée, et système de prévention d'accident d'embarquement et de débarquement l'utilisant Ceased WO2017078213A1 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
KR1020150154989A KR101697520B1 (ko) 2015-11-05 2015-11-05 촬영영상에서의 움직임 객체 검출 방법 및 이를 이용한 차량 승하차 사고 예방 시스템
KR10-2015-0154989 2015-11-05

Publications (1)

Publication Number Publication Date
WO2017078213A1 true WO2017078213A1 (fr) 2017-05-11

Family

ID=58151700

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/KR2015/013119 Ceased WO2017078213A1 (fr) 2015-11-05 2015-12-03 Procédé pour détecter un objet en déplacement dans une image photographiée, et système de prévention d'accident d'embarquement et de débarquement l'utilisant

Country Status (2)

Country Link
KR (1) KR101697520B1 (fr)
WO (1) WO2017078213A1 (fr)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107316275A (zh) * 2017-06-08 2017-11-03 宁波永新光学股份有限公司 一种光流辅助的大尺度显微图像拼接算法
GB2572006A (en) * 2018-03-16 2019-09-18 Continental Automotive Gmbh Vehicle alighting assistance device
WO2019190171A1 (fr) * 2018-03-30 2019-10-03 삼성전자주식회사 Dispositif électronique et procédé de commande associé
CN110723096A (zh) * 2018-07-17 2020-01-24 通用汽车环球科技运作有限责任公司 用于检测被夹柔性材料的系统及方法
CN111401114A (zh) * 2019-02-12 2020-07-10 深圳市艾为智能有限公司 一种基于受限光流场的横穿物体检测装置及方法

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR102471006B1 (ko) * 2018-01-23 2022-11-25 현대자동차주식회사 카메라를 활용한 차량 도어부 승객 안전 장치 및 방법
KR102766407B1 (ko) * 2018-10-01 2025-02-13 삼성전자주식회사 냉장고, 서버 및 냉장고의 객체 인식 방법
US11893704B2 (en) 2019-02-21 2024-02-06 Korea Advanced Institute Of Science And Technology Image processing method and device therefor
CN110598668A (zh) * 2019-09-20 2019-12-20 深圳市豪恩汽车电子装备股份有限公司 机动车盲区检测方法及系统
KR20250032164A (ko) 2023-08-30 2025-03-07 대신네트웍스춘천(주) 인공지능 기반의 사고차량 감지 시스템

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20050096865A (ko) * 2005-08-24 2005-10-06 한민홍 버스 승객하차 안전 시스템
KR20080073933A (ko) * 2007-02-07 2008-08-12 삼성전자주식회사 객체 트래킹 방법 및 장치, 그리고 객체 포즈 정보 산출방법 및 장치
KR20140108828A (ko) * 2013-02-28 2014-09-15 한국전자통신연구원 카메라 트래킹 장치 및 방법
KR101487165B1 (ko) * 2013-07-26 2015-01-28 아진산업(주) 차량 승객을 위한 승하차 안전 시스템

Family Cites Families (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2001239832A (ja) * 2000-02-29 2001-09-04 Matsushita Electric Ind Co Ltd 車両のドア開放時のための警告装置
KR101283965B1 (ko) * 2010-12-23 2013-07-09 전자부품연구원 적응적 사고 검출 시스템 및 사고 검출 방법
KR101270602B1 (ko) 2011-05-23 2013-06-03 아진산업(주) 차량 주변 영상 제공 방법
JP5442701B2 (ja) 2011-11-28 2014-03-12 シャープ株式会社 画像処理方法、画像処理装置およびそれを備えた画像形成装置、画像読取装置、プログラム並びに記録媒体

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
KR20050096865A (ko) * 2005-08-24 2005-10-06 한민홍 버스 승객하차 안전 시스템
KR20080073933A (ko) * 2007-02-07 2008-08-12 삼성전자주식회사 객체 트래킹 방법 및 장치, 그리고 객체 포즈 정보 산출방법 및 장치
KR20140108828A (ko) * 2013-02-28 2014-09-15 한국전자통신연구원 카메라 트래킹 장치 및 방법
KR101487165B1 (ko) * 2013-07-26 2015-01-28 아진산업(주) 차량 승객을 위한 승하차 안전 시스템

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
PARK, S. R. ET AL.: "Development of a Boarding Safety Assistance System Using Moving Object Detection Algorithm", PRODEEDINGS OF KSPE 2015 SPRING CONFERENCE, 15 May 2015 (2015-05-15) *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107316275A (zh) * 2017-06-08 2017-11-03 宁波永新光学股份有限公司 一种光流辅助的大尺度显微图像拼接算法
GB2572006A (en) * 2018-03-16 2019-09-18 Continental Automotive Gmbh Vehicle alighting assistance device
WO2019190171A1 (fr) * 2018-03-30 2019-10-03 삼성전자주식회사 Dispositif électronique et procédé de commande associé
US11430137B2 (en) 2018-03-30 2022-08-30 Samsung Electronics Co., Ltd. Electronic device and control method therefor
CN110723096A (zh) * 2018-07-17 2020-01-24 通用汽车环球科技运作有限责任公司 用于检测被夹柔性材料的系统及方法
CN110723096B (zh) * 2018-07-17 2022-05-03 通用汽车环球科技运作有限责任公司 用于检测被夹柔性材料的系统及方法
CN111401114A (zh) * 2019-02-12 2020-07-10 深圳市艾为智能有限公司 一种基于受限光流场的横穿物体检测装置及方法
CN111401114B (zh) * 2019-02-12 2023-09-19 深圳市艾为智能有限公司 一种基于受限光流场横穿物体检测装置的横穿物体检测方法

Also Published As

Publication number Publication date
KR101697520B1 (ko) 2017-02-02

Similar Documents

Publication Publication Date Title
WO2017078213A1 (fr) Procédé pour détecter un objet en déplacement dans une image photographiée, et système de prévention d'accident d'embarquement et de débarquement l'utilisant
US9785842B2 (en) Safety alarm system and method for vehicle
US20090128632A1 (en) Camera and image processor
JP2009143722A (ja) 人物追跡装置、人物追跡方法及び人物追跡プログラム
WO2018098915A1 (fr) Procédé de commande d'un robot de nettoyage et robot de nettoyage
WO2017065347A1 (fr) Drone de protection et système de protection mobile l'utilisant
CN106926794B (zh) 车辆监控系统及其方法
WO2014051337A1 (fr) Appareil et procédé pour détecter un événement à partir d'une pluralité d'images photographiées
JP2014055925A (ja) 画像処理装置、オブジェクト検出方法、およびオブジェクト検出プログラム
WO2021118047A1 (fr) Procédé et appareil pour évaluer une responsabilité d'accident dans une image d'accident en utilisant l'apprentissage profond
KR102570973B1 (ko) 무인변전소 감시 시스템 및 그 동작 방법
JP6046283B1 (ja) エレベータのドアシステム
KR101219407B1 (ko) 영상처리기술을 이용한 에스컬레이터 모니터링 시스템
KR20160069685A (ko) 주차차단기의 번호인식시스템
WO2013180450A1 (fr) Système de détection de débarquement sécurisé de passagers par utilisation d'une caméra et son procédé
JP6295191B2 (ja) エレベータ乗りかごの画像監視装置
JP2014084064A (ja) プラットホームドア監視装置及びプラットホームドア監視方法
WO2018097384A1 (fr) Appareil et procédé de notification de fréquentation
JP2013052738A (ja) 駆け込み乗車検知装置
WO2021172833A1 (fr) Dispositif de reconnaissance d'objets, procédé de reconnaissance d'objets et support d'enregistrement lisible par ordinateur pour le mettre en œuvre
KR101266015B1 (ko) 승강장 안전선 및 선로 감시 시스템
WO2020218716A1 (fr) Dispositif de stationnement automatique et procédé de stationnement automatique
KR102063957B1 (ko) 지게차 작업사고를 방지하는 방법 및 시스템
WO2018101605A1 (fr) Appareil de surveillance d'image pour afficher des informations d'événement
KR101470939B1 (ko) 감시카메라와 이를 이용한 얼굴 검출 시스템 및 얼굴 검출 방법

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15907876

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15907876

Country of ref document: EP

Kind code of ref document: A1