KR102351476B1 - Apparatus, system and method for object collision prediction - Google Patents

Abstract

Disclosed is an image analysis apparatus for predicting object collision by using a road image. According to the present invention, the image analysis device comprises: an object detection unit detecting a representative trajectory of an entry object entering a road from object collision risk information including a representative trajectory acquired on the basis of a movement trajectory of an object passing through the road; a trajectory analysis unit selecting a candidate representative collision trajectory intersecting the representative trajectory from the object collision risk information and determining whether a candidate collision object exists on the candidate representative collision trajectory; a required time calculation unit calculating a time required for the entry object and the candidate collision object to reach an intersection point of the representative trajectory and the candidate representative collision trajectory when it is determined that the candidate collision object exists as a determination result of the trajectory analysis unit; and an event generation unit determining that a collision prediction event has occurred when a difference from the required time is less than a preset time.

Description

Apparatus, system and method for object collision prediction

The present invention relates to an apparatus, system and method for predicting object collision, and more particularly, detects and tracks a moving object in an image photographed of a road or passage, and uses object collision risk information to collide with the moving object. It relates to detecting a likely collision candidate moving object.

In addition, it relates to a method for constructing or generating the object collision risk information.

Conventionally, the occurrence of a traffic accident in an intersection where vehicles, pedestrians, etc. cross and pass is caused by the operator seeing the screen image of the control center and stopping the vehicle or pedestrian at the intersection due to the occurrence of a collision accident, or a person who gets out of a stationary vehicle After checking the video of the appearance of the accident, it was possible to recognize the accident. Since the confirmation of such a traffic accident corresponds to the recognition after the traffic accident has occurred, it was far from active and active monitoring or control of the traffic accident, and accordingly, it took a long time to process after the traffic accident.

The present invention is to improve the recognition or recognition of collisions in the intersection described above, and in more detail, to propose an apparatus, system, or method capable of predicting the occurrence of collisions on the road, and to achieve it through a solution to be described later do.

In addition, the present invention intends to propose information (or map) capable of indicating the degree of collision risk in the road, or a method for constructing or generating the same, and achieves it through a solution to be described later.

The problems to be solved to be achieved in the present invention are not limited to the problems to be solved, and other problems not mentioned may be clearly understood by those of ordinary skill in the art to which the present invention belongs from the description below. There will be.

According to the present invention, an image analysis apparatus for predicting object collision using a road image is proposed, wherein the image analysis apparatus is obtained from object collision risk information including a representative trajectory obtained based on a movement trajectory of an object passing on a road. an object detection unit detecting a representative trajectory of an entry object entering the road; a trajectory analysis unit that selects a representative trajectory of a collision candidate that intersects the representative trajectory from the object collision risk information and determines whether a collision candidate object exists on the representative trajectory of the collision candidate; a time required calculation unit for calculating the time required for the entry object and the collision candidate object to reach an intersection of the representative trajectory and the representative trajectory of the collision candidate, respectively, when the collision candidate object exists as a result of the determination of the trajectory analysis unit; and an event generator that determines that a collision prediction event has occurred when the difference between the required time is less than a preset time.

Additionally or alternatively, the object collision risk information includes a plurality of representative trajectories generated by accumulating and recording movement trajectories of objects passing through the road, grouping them, and information on intersections between the plurality of representative trajectories. can do.

Additionally or alternatively, the object collision risk information may include information on a collision risk associated with an intersection between the plurality of representative trajectories.

Additionally or alternatively, the information on intersections between the plurality of representative trajectories includes: a number of representative trajectories crossing the intersection, an average crossing time interval between representative trajectories crossing the intersection, and at each of the representative trajectories. It may include at least one of an average moving speed to the intersection point or a distance to the intersection point.

Additionally or alternatively, the object collision risk is: a greater number of candidate representative trajectories crossing the intersection indicates a higher risk, a shorter average crossing time interval indicates a greater risk, or the average moving speed to the intersection is A higher risk or a shorter distance to the crossing point may indicate a higher risk.

Additionally or alternatively, if there are two or more candidate representative trajectories intersecting the representative trajectory, one of the two or more candidate representative trajectories may be selected according to the object collision risk associated with the intersection according to the object collision risk information.

Additionally or alternatively, a time required for the object and the collision candidate object to reach the intersection is calculated according to an average movement speed of the object and the collision candidate object, wherein the average movement speed is a movement per unit time of each object It can be based on the number of pixels or can be measured with a radar sensor.

Additionally or alternatively, the preset time may be variably set according to the object collision risk associated with the intersection according to the object collision risk information.

Additionally or alternatively, the image analysis apparatus may further include an image acquisition unit configured to acquire the image.

Additionally or alternatively, the image analysis apparatus may further include a radar sensor for detecting and tracking the object.

The image analysis apparatus may further include an operation unit for receiving the image, storing or displaying the image, or receiving and displaying the collision prediction event.

The image analysis apparatus may further include an operation unit that receives information about an object from the object detection unit and generates and displays at least one of traffic information, illegal parking information, speed violation information, and life crime prevention information.

The above problem solving methods are only some of the embodiments of the present invention, and various embodiments in which the technical features of the present invention are reflected are based on the detailed description of the present invention to be described below by those of ordinary skill in the art can be derived and understood as

According to the present invention, it is possible to predict the collision between objects in the intersection and provide the prediction information to the control or operation center, so that the operator can recognize the occurrence of an accident in advance, thereby inducing faster post-accident processing in time. can

According to the present invention, it is possible to calculate the object collision risk in an intersection, and by cumulatively managing it, it is possible to increase the accuracy of object collision prediction.

The effects obtainable in the present invention are not limited to the above-mentioned effects, and other effects not mentioned may be clearly understood by those of ordinary skill in the art to which the present invention belongs from the following description. will be.

BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings, which are included as a part of the detailed description for better understanding of the present invention, provide embodiments of the present invention, and together with the detailed description, explain the technical spirit of the present invention.
1 shows the overall configuration of a traffic control system related to the present invention.
2 is a diagram illustrating object collision risk information according to the present invention.
3 is for explaining a procedure for constructing or generating object collision risk information according to the present invention.
4 shows a flowchart of a method for object collision prediction according to the present invention.
5 is for explaining a process of calculating a time difference between arrival/arrival at an intersection point or a collision prediction point of an object moving along two representative trajectories according to the present invention.
6 shows the configuration of an apparatus for object collision prediction according to the present invention.

Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. However, the present invention is not limited to the embodiments described herein and may be implemented in various other forms. The terminology used herein is for the purpose of helping the understanding of the embodiments, and is not intended to limit the scope of the present invention. Also, singular forms used hereinafter include plural forms unless the phrases clearly indicate the opposite.

1 shows the overall configuration of a traffic control system in the field to which the present invention pertains. In the system, an image acquisition device 1 such as a camera for acquiring an image by photographing a traffic situation and related accessories, a server 2 connected to the image acquisition device by wire or wirelessly, and the image It may include a control system (3) such as a traffic control room operated by various local governments, road corporations, facilities corporations, and the National Police Agency, etc. that receive and control real-time or store and manage images.

The device, system, etc. according to the present invention may be included in any one of the components shown in FIG. 1 or may exist as a separate device, system, or the like. That is, the system related to the present invention includes an image acquisition device (or an image acquisition unit) such as a camera for acquiring an image by photographing a traffic situation, associated devices, and a wired connection with the image acquisition device (or image acquisition unit) Alternatively, a server connected wirelessly, and a control system such as a traffic control room operated by various local governments, road corporations, facilities corporation, and the National Police Agency that receive the video and control it in real time or store and manage the video ( or operation department). The image acquisition device (or image acquisition unit) may operate in conjunction with means for detecting a vehicle such as a vehicle detection sensor (eg, at least one of an electromagnetic sensor, a radar sensor, and an image sensor).

2 illustrates a process for explaining object collision risk information according to the present invention.

Although expressed as “object collision” in this specification, examples thereof include a vehicle-to-vehicle, vehicle-to-person traffic accident, or a human-to-person collision.

One or more intersection images may be acquired through equipment such as a camera for photographing the intersection (a). In the sequentially acquired intersection images, moving objects (eg, vehicles, hereinafter simply referred to as “objects”) may be detected and their movements tracked, and the movement trajectories of the objects within the intersection may be accumulated and recorded (b) . Until it is possible to group the movement trajectories of objects, the accumulation and recording may continue. Then, the movement trajectories of the objects may be grouped (c), and the grouped movement trajectories may be referred to as representative trajectories. For example, each representative trajectory represents a trajectory corresponding to movement (eg, going straight, turning right, turning left, etc.) allowed for each lane when lanes are divided on the road. In FIG. 2C , the representative trajectories are indicated by straight lines, curves, etc. having different colors for each intersection entry position. Some of the representative trajectories form an intersection point overlapping with other representative trajectories, and the greater the number of representative trajectories forming the intersection point, the higher the probability of a collision or accident is expected. It can be expressed for each intersection as In FIG. 2d , the higher the risk, the larger the red circle. That is, the object collision risk indicates the number of representative trajectories forming each intersection, and the greater the number of intersecting representative trajectories, the higher the object collision risk.

Also, the object collision risk may be expressed not only as the number of intersecting representative trajectories, but also as a (average) time difference or time interval at which two representative trajectories forming an intersection intersect. Even if the representative trajectories obtained through the accumulation and grouping of the movement trajectories described above form an intersection with each other, in general, objects on the two representative trajectories will not pass through the intersection at the same time, except in the case of an accident or collision. Thus, each representative trajectory is unlikely to actually intersect, but may have a “time difference or time interval” to reach that intersection.

For example, the (average) time difference or time interval at which the two representative trajectories intersect is, assuming that the representative trajectory A and the representative trajectory B form an intersection point C, the object a moving through the representative trajectory A is the intersection point Corresponds to the difference or interval between the time to reach C and the time for an object b moving through the representative trajectory B to reach the intersection point C.

Referring to FIG. 5 , assuming that the cycle of traffic light permitting traffic operates in the order of ramps 1, 2, 3, and 4, the representative trajectory of starting (entrying) from ramp No. 1 and departure (entry) from ramp No. 4 ) passing through the intersection (i.e., intersection) of the representative trajectories (indicated by the “collision prediction interval” in Fig. 5), the object moving along each representative trajectory is, on average, for the object of the ramp within the period of the corresponding traffic light. It will pass through that intersection with a time difference or time interval for which a proceeding permit signal (green light, left turn signal, etc.) is illuminated. This is because it is obtained as a result of accumulating and recording the trajectories of numerous objects. In this example, the time difference or time interval is 1/4 section (vehicle entering No. 4 -> vehicle entering No. 1) or 3/4 section (with No. 1) of the entire cycle, considering the cycle of the traffic light. It will correspond to the vehicle entering -> the vehicle entering No. 4).

In FIG. 5 , from the point of view of the object departing (entering) from the first ramp, according to the sequence assumption of the traffic light operation, rather than the object departing (entering) from the fourth ramp, the object departing (entering) from the third ramp The probability of collision is high. Accordingly, the object collision risk may be expressed as a (average) time difference or time interval at which two representative trajectories intersect.

Also, the object collision risk may represent a distance from an intersection entry point of a corresponding representative trajectory to a point where it intersects with another representative trajectory, which may be expressed by the number of pixels. Alternatively, the object collision risk may indicate a distance from an intersection entry point of another representative trajectory b intersecting the corresponding representative trajectory a to an intersection point of the two representative trajectories a and b. That is, when two representative trajectories intersect, the information on the distance related to the corresponding intersection may include information on the distance from the entry point of each representative trajectory to the intersection. In addition, information about the distance from the entry point of each representative trajectory to the intersection point as well as the distance from all points of each representative trajectory to the intersection point may be included, and preferably, the corresponding intersection point according to the movement direction on each representative trajectory For points after passing through , information on distance may be excluded. At this time, the risk indicated by the object collision risk may be higher as the distance is shorter, that is, the closer the distance from the intersection entry part or each point on the representative trajectory (for example, the current location of the object) to the intersection point with other representative trajectories. have.

Alternatively, the object collision risk is the average moving speed (v_c) from the point at which the representative trajectory (c) intersects with another representative trajectory (d) from the intersection entry point, or the object moving along the representative trajectory (c). It may represent an average moving speed (v_c'). Alternatively, the object collision risk is the average moving speed (v_d) from the intersection entry point to the point where another representative trajectory (d) intersecting the representative trajectory (c) intersects the representative trajectory (c), or the representative trajectory It may represent the average moving speed (v_d') of the object moving along (d). In this case, the risk indicated by the object collision risk may be higher as the moving speed of the corresponding representative trajectory increases.

Accordingly, the object collision risk information includes information related to the representative trajectory(s) or an intersection formed by at least two representative trajectories, and the information related to the intersection may include the object collision risk described above representatively. As will be described later, the object collision risk is related to a collision candidate object (or a representative collision candidate trajectory) highly related to the first object for object collision prediction for each object entering the intersection (referred to as a “first object”). It is used to determine object collision prediction in the order of the lowest collision candidate object (or collision candidate representative trajectory).

In addition, each representative trajectory included in the object collision risk information may be managed by being assigned an identification number. In addition, the object collision risk information includes intersection information between respective representative trajectories, that is, for each representative trajectory, information on other representative trajectory(s) intersecting the representative trajectory may be stored in advance. Corresponding information may be identified and stored in a procedure of constructing or generating object collision risk information related to FIG. 3 to be described later.

Meanwhile, in FIGS. 1 and 2 , the content related to the present invention is limited to a vehicle, but the present invention may be applied to a vehicle, a pedestrian, and all other movable objects.

3 is for explaining a procedure for constructing or generating object collision risk information according to the present invention. In the description related to FIG. 3 below, it is referred to as “object collision risk information”, but the scope of the present invention is not limited by the name. Accordingly, the object collision risk information may be expressed as an object collision risk map or information related to an object collision risk for a corresponding road or passage, or an intersection of a road or passage.

Although the subject performing the procedure according to FIG. 3 is not specified, the image analysis apparatus 60 for object collision prediction according to the present invention, or a specific component in the image analysis apparatus, or an information generating unit for generating object collision risk information and the like, and the name or name of the component does not limit the scope of rights of the corresponding content.

An image of a road is input to the device 60 (S31). The device 60 may detect an object from the input image (S32). When an object is detected, the device 60 may track the detected object ( S33 ), record the path the object moves, and store it as one trajectory. The device 60 may determine whether tracking of the object is completed (S34). A condition for determining whether tracking of the object is completed may be preset, and this may be set as the number of images (frames) or a time period. When the tracking is completed, the device 60 may accumulate and record the trajectories stored up to this point (S35). Then, the device 60 may determine whether the trajectories can be grouped (clustered) based on all trajectories that are accumulated and recorded (S36). For example, it may be determined whether the number of accumulated trajectories or whether grouping of the accumulated trajectories is possible in consideration of mutual deviation or correlation. The grouping means generating one representative trajectory from a plurality of overlapping trajectories. If the device 60 detects and tracks an object for a number of input images, trajectories will be accumulated in proportion to the number of input images, and accordingly, trajectories having the same movement line can be represented by one representative trajectory, , you can save it.

As a result of the grouping, the object collision risk information may be generated, and information related to the representative trajectory(s) S37-1 and the intersection (or intersection) between the two representative trajectories may be included. The information related to the crossing point may include a crossing time interval between representative trajectories (S37-2), a distance to the crossing point, or an average moving speed (S37-3).

Here, the information related to the intersection point may be information provided for each representative trajectory. Specifically, for the first representative trajectory, a time interval or intersection between the first representative trajectory and one or more other representative trajectories that intersect (overlapping) the first representative trajectory. time interval; and an average moving speed of the object on the first representative trajectory up to an intersection point between one or more other representative trajectories intersecting (overlapping) the first representative trajectory.

In addition, the "time interval or intersecting time interval between representative trajectories" refers to the accumulated and recorded average time difference. It means an average difference between the time passing through the intersection of the moving object and the first representative trajectory. Since the time difference includes the forward direction and the opposite direction according to the passage of time, it is defined as a value for one of these two directions, and preferably means the difference according to the (forward) flow of time.

The object collision risk information may be periodically updated, and a plurality of trajectories that are the basis of the object collision risk information may be accumulated and managed. Since a larger number of parent data can guarantee higher accuracy, the accuracy of object collision prediction can also be increased by updating the object collision risk information and accumulative management of a plurality of trajectories.

4 is a flowchart of a method for object collision prediction using a road image according to the present invention. Although the subject of performing the procedure according to FIG. 4 is not specified, it may be performed by the image analysis apparatus for object collision prediction according to the present invention, or a specific component in the image analysis apparatus, and the name or name of the component is It does not limit the scope of rights of the content.

The device may receive an image of the road (S411). Since it is for collision prediction, the image should be inputted simultaneously with or immediately after shooting. The device may detect an object from the input image (S412) and track the object (S413). When the object is detected and tracked, the device may detect a representative trajectory related to the object from the previously generated object collision risk map (S414).

Then, the device may select another representative trajectory that intersects the detected representative trajectory (referred to as a “collision candidate representative trajectory”) (S415). This is to detect an object on the representative trajectory of all collision candidates that may collide with the object. Representative trajectories that do not intersect the selected representative trajectory on which the object is located are not considered in the selection because it is determined that there is no object collision possibility.

However, the present invention relates to object collision prediction, and it is preferable that the object collision prediction is performed first for a portion with a high risk of collision occurrence. Accordingly, when there are several representative trajectories of collision candidates intersecting the selected representative trajectory, it is preferable to determine the order of the representative trajectories of the collision candidates for which object collision prediction is to be performed among the plurality of representative trajectories. Accordingly, the object collision risk information described above can be used to determine the order of the collision candidate representative trajectories.

That is, in selecting the collision candidate representative trajectory that intersects the detected representative trajectory, the device may use the object collision risk information.

When the representative collision candidate trajectory is selected, the device may determine whether another object exists on the corresponding collision candidate representative trajectory or whether there is another object to be performed a procedure to be described later (S416). If there is no other object on the representative trajectory of the collision candidate or there is no longer an object to perform procedures such as S417 to S420 to be described later on the representative trajectory of the collision candidate, the device determines whether there are other intersecting representative trajectories of the collision candidate It can be determined (S421). If there is another object on the corresponding collision candidate representative trajectory and the corresponding object is an unspecified collision candidate object, the device may designate the corresponding object as a collision candidate object ( S417 ). Then, the device may calculate the time required to reach the intersection (intersection) of the two representative trajectories of the two objects (S418). If there is another object on the corresponding collision candidate representative trajectory and the corresponding object is a predefined collision candidate object, the apparatus may perform S418 without designating the collision candidate object.

The device may determine whether the required time to the intersection of the two objects is less than a preset time (S419). If the required time is less than a preset time, the device may generate a collision prediction event (S420). This means that since it is determined that the two objects arrive at the intersection point close in time, the probability of collision is high. Accordingly, the device generates a collision prediction event. Then, it may be determined whether another object exists on the representative trajectory of the corresponding collision candidate ( S416 ). (That is, if there are a plurality of objects on the corresponding collision candidate representative trajectory, since S417 to S420 must be performed for another object)

On the other hand, if the required time is equal to or longer than the preset time, the device may determine whether another object is present on the representative trajectory of the corresponding collision candidate ( S416 ), since it means that the probability of occurrence of a collision between the two objects is low. (That is, if there are a plurality of objects on the corresponding collision candidate representative trajectory, since S417 to S420 must be performed for another object)

If the above procedure is repeated, the operation for object collision prediction with respect to the representative trajectory of the collision candidate selected in S415 is completed.

Then, the device may determine whether there are other representative trajectories of collision candidates (S421).

If there are remaining representative trajectories of collision candidates in the corresponding input image, the apparatus may proceed to S415 to select the remaining representative trajectories of collision candidates (if necessary, according to priority); Otherwise, it may be determined whether the above procedure has been completed for all objects of the corresponding image (S422). If there is an object that has not been performed, the procedure returns to S412, and the device starts detecting and tracking another object. If the procedure for the representative trajectories of collision candidates for all objects in the image is completed, the procedure may be set to receive the next image, or if there is a preset condition, the procedure may be terminated.

Although the procedure according to FIG. 4 is shown to be continuously performed when images exist indefinitely, it may be performed and terminated for only one input image, or may be performed and terminated for a certain number of input images.

5 is for explaining a process of calculating a time difference between arrival/arrival at an intersection point or a collision prediction point of an object moving along two representative trajectories according to the present invention. Numbers shown in FIG. 5 indicate an access road entering the corresponding intersection at the crossroads intersection. For convenience, it is assumed that traffic flow occurs in the order of ramps 1, 2, 3, and 4 under normal traffic light control (ie, 1->2->3->4->1->2->). order of ..).

Assuming that it is detected that the first object enters through the first access road in the corresponding image, the representative trajectory of the first object may be selected as the second representative trajectory. Then, one or more remaining representative trajectories that intersect the second representative trajectory may be searched using the object collision risk map, and one of them is the first representative trajectory. Assume it exists The intersection of the first representative trajectory and the second representative trajectory becomes a point marked as a “collision prediction section”, which means that when an object moves along the two representative trajectories, an intersection, that is, a collision or an accident, is possible. However, collision prediction is not possible only by this, and it can be determined that an accident is possible when the arrival times of the two objects to the corresponding "collision prediction section" are the same or very close. Accordingly, it is necessary to calculate the time required for the object moving on each of the two representative trajectories to reach the collision prediction section, and information included in the object collision risk information may be used to calculate the corresponding time. For example, "average moving speed" or "distance" from the current location to the intersection ("collision prediction interval") may be used in the time calculation.

6 shows the configuration of the image analysis apparatus 60 for object collision prediction according to the present invention. The device includes an information generating unit 61 , an object detecting unit 62 , a trajectory analyzing unit 63 , a required time calculating unit 64 , an event generating unit 65 , a control unit 66 , a communication unit 67 and a storage unit. (68) and the like.

The information generating unit 61 may construct or generate object collision risk information for the intersection of the corresponding road or passage described with reference to FIG. 3 .

The object collision risk information may include representative trajectory(s) obtained based on a movement trajectory of an object passing on a road in a road image. In addition, the object collision risk information may include information related to an intersection (or intersection) between two representative trajectories. The information related to the crossing point may include a crossing time interval between representative trajectories, a distance to the crossing point, or an average moving speed. For detailed information on the object collision risk related information, refer to the description related to FIG. 3 .

The information generating unit 61 may update the object collision risk information periodically or according to a request of the control system or a device belonging thereto, or a user controlling the same. This has the effect of automatically resolving detection errors by updating the map when a region of interest (RoI) is changed due to a shift in the camera angle of view.

When the road image is input, the object detector 62 may detect an object and track the detected object. An object detected and tracked by the object detection unit 62 is referred to as a first object for convenience of description. Then, the object detection unit 62 may detect the representative trajectory of the entry object entering the road from the image from the object collision risk information. The representative trajectory of the first object is a trajectory in which the first object is expected to move within the intersection according to the detection and tracking. The representative trajectory for the first object may or may not be changed whenever a road image is input. For example, the representative trajectory may be changed when the first object changes a lane, a lane, or a passage in an intersection or makes an equivalent motion.

The object detection unit 62 may distinguish and detect and recognize objects such as people, vehicles, and license plates. This is to be used for various purposes, and specifically, the object detection unit 62 is a traffic information collection system, an illegal parking and stopping enforcement system, an unmanned road crime prevention system, a living crime prevention system, and an intelligent It can operate in conjunction with CCTV systems classified as object tracking systems, unmanned signal/speed enforcement systems, and parking management systems. To this end, the object detection unit 62 detects and recognizes an object necessary for traffic information collection, illegal parking, speed violation or life crime prevention, and provides information on the detected and recognized object to various local governments, road corporations, facilities corporation, It can be provided by a control system (or operation department) such as a traffic control room operated by the National Police Agency.

The trajectory analysis unit 63 may select one or more collision candidate representative trajectories that intersect the representative trajectory from the object collision risk information. When a representative trajectory of the first object (referred to as a “first representative trajectory”) is selected, information on all representative trajectories of collision candidates intersecting the first representative trajectory may be detected, or the object collision risk information Information on all representative trajectories of collision candidates that intersect the first representative trajectory may be stored in advance. Information on such intersecting collision candidate representative trajectories may be previously stored for all representative trajectories.

When there are a plurality of the representative trajectories of the collision candidates, it is preferable to determine a priority among the representative trajectories of the collision candidates, and for this purpose, the object collision risk information may be used. An operation according to the present invention is performed on the collision candidate object on the corresponding trajectory from the high-priority collision candidate representative trajectory.

The object collision risk information includes information on the object collision risk associated with an intersection between the plurality of representative trajectories. The representative trajectories include the collision candidate representative trajectories. The information on the intersection between the plurality of representative trajectories includes at least the number of representative trajectories crossing the intersection, the average crossing time interval of the representative trajectories crossing the intersection, and the average moving speed to the intersection or the distance to the intersection. may contain one.

The object collision risk indicates a higher risk as the number of representative trajectories crossing the intersection between the plurality of representative trajectories increases, and the shorter the average crossing time interval of the representative trajectories crossing the intersection, the higher the risk, or up to the intersection The higher the average moving speed or the shorter the distance to the intersection point, the higher the risk may be, and the higher the risk, the higher the priority as described above.

When there are two or more collision candidate representative trajectories intersecting the representative trajectory, a collision candidate object is detected from the collision candidate representative trajectory having a high object collision risk according to the object collision risk information (if present), and the representative trajectory and A time required to an intersection (“intersection”) between the corresponding collision candidate representative trajectories may be calculated. The trajectory analysis unit 63 may determine whether there is a collision candidate object entering or moving at the intersection through the representative collision candidate trajectory.

If there is no collision candidate object, that is, if there is no object related to all collision candidate representative trajectories or if the procedure to be described later (also described above) is performed on the objects related to all collision candidate representative trajectories, all operations on the corresponding input image are completed. Then, if there is an input image, object detection and tracking are started from the beginning again.

When there is the collision candidate object, the required time calculation unit 64 may calculate a time required for the first object and the collision candidate object to reach an intersection. Alternatively, the required time calculation unit 64 may calculate the distance to the intersection (the intersection between the first representative trajectory and the collision candidate representative trajectory).

The time required for the first object and the collision candidate object to reach the intersection is calculated according to the average movement speed of the first object and the collision candidate object, and the average movement speed is the movement per unit time of the object in the image. It can be based on the number of pixels or can be measured with a radar sensor. As described above, the object collision risk information includes the average moving speed or distance to all intersections for each representative trajectory, so The time required to the expected intersection can be calculated using

When the difference between the time required for the first object and the collision candidate object to reach the intersection is less than a preset time, the event generator 65 may determine that a collision prediction event has occurred. Otherwise, it is determined that the collision prediction event has not occurred. Meanwhile, the preset time may be variably set according to the object collision risk according to the object collision risk information. For example, the preset time may be set to be relatively small as the risk of object collision with respect to the representative trajectory of the collision candidate or the intersection point is high.

Thereafter, as described with reference to FIG. 4 , the procedure described above may be repeated for the next collision candidate object, or the next priority collision candidate representative trajectory, or the next object in the image, or the next input image. . Also, when it is determined that the collision prediction event has occurred, the device 60 may transmit the corresponding information to another device or system through the communication unit 67 . If the information is received from a control system such as a traffic control room, the real-time image of the intersection may be enlarged and displayed on the display device, or a visual or audible alarm may be generated through an alarm display device such as a warning light or a speaker. can occur.

The control unit 66 includes the information generating unit 61 , the object detecting unit 62 , the trajectory analyzing unit 63 , the required time calculating unit 64 , the event generating unit 65 , and the communication unit 67 . ), to control the operation of the storage unit 68 or to instruct or perform an operation related to the present invention. In addition, the control unit 66 may perform a function of storing or classifying, processing, or providing information generated by each of the components. For example, although not described with reference to FIG. 6 , the control unit 66 performs an operation on at least a portion of S416 , S421 , and S422 of FIG. 4 , and accordingly the object detection unit 62 and the trajectory analysis unit (63), etc., may instruct the next operation.

In addition, the image analysis apparatus 60 may further include an image acquisition unit (not shown) for acquiring the image, and input an image from the image acquisition unit for acquiring the image present separately from the device. can receive The image acquisition unit may include a radar sensor for detecting and tracking the object. In addition, the image analysis apparatus 60 may further include an operation unit that receives and stores or displays the image or receives and displays the collision prediction event. The operation unit may receive information on the detected and recognized object from the object detection unit 62 to generate and display traffic information, illegal parking information, speed violation information, or life crime prevention information.

Meanwhile, all contents of the present invention described with reference to FIG. 4 may be applied to the device illustrated in FIG. 6 . In addition, all of the components 61 to 68 described with reference to FIG. 6 have been illustrated and described as components included in the device 60, but some of them are implemented while being physically separated from the device or a separate device or a separate device. may be included.

In addition, in the previous description, two values are compared as in S419. When comparing two values, an equal sign may be included in the magnitude relationship. That is, in the specification, "less than" instead of "less than", and "more than" instead of "more than" may be applied, and vice versa, which does not limit the scope of the present invention.

In the above specification, it has been described that "device" and its components (object detection unit, trajectory analysis unit, required time calculation unit, etc.) perform the method or procedure, etc., but "device" and components belonging thereto are only names. The scope of rights is not subordinated to it. That is, the method or procedure may be performed not only as a device but also as a system, and the method or method may be performed by software for object collision prediction, or code readable by a computer or other machine, device, etc. .

In addition, as another aspect of the present invention, the above-described proposal or operation of the invention is performed by a "computer" (a comprehensive concept including a system on chip (SoC) or (micro) processor, etc.) Code that can be implemented, implemented or executed, or a computer-readable storage medium storing or including the code, or a computer program product, etc. a computer-readable storage medium including a computer program product or a computer program product.

The detailed description of the preferred embodiments of the present invention disclosed as described above is provided to enable any person skilled in the art to make and practice the present invention. Although the above has been described with reference to preferred embodiments of the present invention, it will be understood by those skilled in the art that various modifications and changes can be made to the present invention described in the claims below. Accordingly, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims (13)

An image analysis apparatus for predicting object collision using a road image, the image analysis apparatus comprising:
an object detection unit detecting a representative trajectory of an entry object entering a road from object collision risk information including a representative trajectory obtained based on a movement trajectory of an object passing through the road;
a trajectory analysis unit that selects a representative trajectory of a collision candidate that intersects the representative trajectory from the object collision risk information and determines whether a collision candidate object exists on the representative trajectory of the collision candidate;
a time required calculating unit for calculating the time required for the entry object and the collision candidate object to reach an intersection point of the representative trajectory and the representative trajectory of the collision candidate, respectively, when the collision candidate object exists as a result of the determination by the trajectory analysis unit; and
When the difference between the required time is less than a preset time, an event generator for determining that a collision prediction event has occurred,
The preset time is variably set according to the object collision risk related to the intersection point according to the object collision risk information. The method of claim 1, wherein the object collision risk information includes information on a plurality of representative trajectories generated by accumulating and recording movement trajectories of objects passing through the road, grouping them, and an intersection point between the plurality of representative trajectories. which is an image analysis device. The image analysis apparatus of claim 2 , wherein the object collision risk information includes information on a collision risk related to an intersection between the plurality of representative trajectories. According to claim 2, wherein the information on the intersection point between the plurality of representative trajectories:
At least one of the number of representative trajectories crossing the intersection, the average crossing time interval between the representative trajectories crossing the intersection, the average moving speed from each of the representative trajectories to the intersection, or the distance to the intersection , an image analysis device. 5. The method of claim 4, wherein the object collision risk is:
The greater the number of candidate representative trajectories crossing the intersection, the higher the risk,
The shorter the average crossover time interval, the higher the risk, or
The higher the average moving speed to the intersection, or the shorter the distance to the intersection, the higher the risk is, the image analysis apparatus. The method of claim 1, wherein if there are two or more candidate representative trajectories intersecting the representative trajectory,
An image analysis apparatus for selecting one of two or more candidate representative trajectories according to an object collision risk associated with the intersection according to the object collision risk information. The method of claim 1, wherein the time required for the object and the collision candidate object to reach the intersection is calculated according to an average movement speed of the object and the collision candidate object, and the average movement speed is a movement per unit time of each object. An image analysis device, either based on the number of pixels or measured by a radar sensor. delete The image analysis apparatus of claim 1, further comprising an image acquisition unit configured to acquire the image. The method of claim 1,
The image analysis apparatus further comprising a radar (radar) sensor for detecting and tracking the object. The image analysis apparatus of claim 1, further comprising an operation unit that receives and stores or displays the image or receives and displays the collision prediction event. The image analysis apparatus of claim 1 , further comprising an operation unit that receives information on an object from the object detection unit and generates and displays at least one of traffic information, illegal parking information, speed violation information, and life crime prevention information. As a method of performing object collision prediction using a road image,
detecting a representative trajectory of an entry object entering a road from object collision risk information including a representative trajectory obtained based on a movement trajectory of an object passing through the road;
selecting a collision candidate representative trajectory that intersects the representative trajectory from the object collision risk information;
determining whether a collision candidate object exists on the collision candidate representative trajectory;
If the collision candidate object exists as a result of the determination in the step of determining whether the collision candidate object exists, it takes the entry object and the collision candidate object to reach the intersection of the representative trajectory and the collision candidate representative trajectory, respectively counting time; and
determining that a collision prediction event has occurred if a difference between the time required for the object and the collision candidate object to reach the intersection is less than a preset time;
The preset time is variably set according to the object collision risk associated with the intersection according to the object collision risk information.

Images