JP2015230579A - Accident image acquisition system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an accident image acquisition system capable of appropriately acquiring an accident image picked up through a vehicle.SOLUTION: An accident image acquisition system comprises: a vehicle 20 transmitting a picked-up image obtained by imaging surroundings of a vehicle; and a center 10 acquiring the picked-up image transmitted from the vehicle 20. The vehicle 20 includes an image recognition unit 25 recognizing an accident image P1 involving an accident vehicle 50 other than an own vehicle from the picked-up image, and transmits the picked-up image recognized as the accident image P1 by the image recognition unit 25 to the center 50. If determining that the accident vehicle 50 and a relative object close to the accident vehicle 50 are present in the picked-up image through image processing on the image, the image recognition unit 25 recognizes whether the picked-up image is the accident image P1 on the basis of a degree of proximity between the accident vehicle 50 and the relative object.

Description

  The present invention relates to an accident image acquisition system that acquires an accident image captured by a vehicle.

  Many recent vehicles detect the occurrence of an accident based on an impact or the like, and can capture the situation of the accident with a camera that captures the front environment that is installed in advance. In addition, a technique is also known in which the center side can recognize the occurrence of an accident involving the vehicle by transmitting a captured image when the accident is detected from the vehicle to the center. Furthermore, in order to grasp the situation of the accident in more detail, a technique has been proposed in which an image relating to the accident occurrence point is obtained from other vehicles traveling around the accident occurrence point. Have been described.

  The system described in Patent Document 1 includes an information processing device provided in an information center and various in-vehicle devices provided in a plurality of vehicles, and the information center collects images around the place where the accident occurred from the plurality of vehicles. That is, the vehicle includes a communication device, a car navigation system, an impact detection unit that detects the occurrence of an accident, a camera that captures the periphery of the vehicle, and a storage unit that stores the captured video in association with the time, And an in-vehicle CPU for controlling them. When an accident is detected in a vehicle, the system transmits information about the accident from the accident vehicle to the information center, so that the information center can report the accident to other vehicles around the location where the accident occurred. Occurrence is notified, and images taken by other vehicles notified are acquired.

JP 2012-98105 A

  According to the system described in Patent Document 1, when an accident occurs, not only the accident vehicle but also an image captured by another vehicle is acquired in the information center. However, even if an accident occurs, an accident with low impact is not detected by the accident vehicle, and there is a possibility that the occurrence of the accident will not be reported in the first place, or that the device for reporting the accident may be broken. In addition, a vehicle not equipped with a device for reporting such an accident may be involved in the accident. That is, the accident image may not be acquired depending on the report from the vehicle that caused the accident.

  This invention is made | formed in view of such a situation, The objective is to provide the accident image acquisition system which can acquire appropriately the accident image imaged through the vehicle.

Hereinafter, means for solving the above-described problems and the effects thereof will be described.
An accident image acquisition system that solves the above problem is an accident image acquisition system that includes a first vehicle that transmits a captured image obtained by imaging the periphery of the vehicle, and a center that acquires a captured image transmitted from the first vehicle. The first vehicle has a recognition unit that recognizes an accident image involving a second vehicle different from the host vehicle from the captured image, and the captured image is recognized as an accident image by the recognition unit. To the center, and the recognizing unit determines that the second vehicle and a relative object close to the second vehicle exist in the image through image processing of the captured image. The gist is to recognize whether or not the captured image is an accident image based on the degree of proximity between the second vehicle and the relative object.

  According to such a configuration, since the captured image recognized as the accident image is transmitted from the first vehicle to the center, the center captures a significant captured image for grasping the situation of the accident involving the second vehicle. Be able to get.

  Even if the second vehicle cannot notify the center of the accident situation, as long as the first vehicle exists in the vicinity of the second vehicle, the accident image captured by the vehicle can be appropriately acquired by the center. . Then, the center can recognize the accident, report the accident, and hold an image significant for grasping the situation of the accident. That is, if an accident is recognized at the center, an image for accident notification, rescue, treatment, and ensuring safety will be provided to the center even if the second vehicle cannot report the accident. become. In the case of failure to report an accident, it may be possible to detect an accident when the device is not equipped with a device for reporting an accident or the device is equipped with a device for reporting an accident but the impact of the accident is weak. For example, when it was not possible or when a device for reporting an accident broke down.

  Furthermore, since the captured image that is recognized as an accident image is acquired at the center, compared to the case where a large amount of non-accident images are included and a necessary image cannot be obtained without screening. The value of the acquired image is naturally increased. Moreover, since only the captured image recognized as the accident image is transmitted, the communication amount can be suppressed.

  Further, the accident image can be appropriately recognized from the captured image by recognizing the accident image based on the degree of proximity between the second vehicle and the relative object close to the second vehicle. .

As a preferred configuration, the first vehicle automatically transmits the recognized accident image to the center in response to recognizing the accident image from the captured image.
According to such a configuration, the above-described operational effects are realized based on the determination of accident recognition in the first vehicle.

As a preferred configuration, the first vehicle transmits position information corresponding to the accident image to the center prior to transmission of the recognized accident image.
According to such a configuration, first, the position information is transmitted, so that it is possible to report the accident quickly and reliably as compared with the transmission of the captured image. In addition, immediately after the accident is recognized, it is possible to suppress a situation in which captured images are transmitted all at once from a plurality of vehicles and the amount of communication increases rapidly.

  As a preferred configuration, the center requests the first vehicle estimated to have captured the accident image based on the received position information to transmit the accident image, and the first vehicle requested to transmit the center receives the accident image. An accident image is transmitted in response to a transmission request from.

  According to such a configuration, the accident image is transmitted from the first vehicle that is estimated to have captured the accident around the accident point, so that the center can acquire and acquire an appropriate accident image. Become. As a result, the center can acquire a necessary accident image while suppressing an increase in the amount of communication with the first vehicle.

  As a preferred configuration, the recognizing unit divides the captured image into a plurality of regions based on pixel characteristics, and identifies a combination in which at least one of the adjacent regions is a “vehicle” from the divided regions. Based on the proximity degree of the two areas corresponding to the specified combination, it is recognized whether or not the image is an accident image.

  According to such a configuration, the process of dividing an image according to features to create a region (super pixel or segmentation), and the process of specifying that a vehicle is included in the created divided region combination Accordingly, it is possible to reduce the load on the image recognition process. And the 2nd vehicle which caused the accident comes to be identified from this combination.

  As a preferred configuration, the recognition unit performs image classification of each region by image processing using a support vector machine, and identifies a combination in which at least one of adjacent regions is “vehicle” based on the image classification.

According to such a configuration, appropriate image classification can be performed by the support vector machine.
As a preferred configuration, the combination includes a combination of an area corresponding to “vehicle” and an area corresponding to “vehicle”, a combination of an area corresponding to “vehicle” and an area corresponding to “motorcycle”, and “vehicle At least one combination of a region corresponding to “bicycle” and a region corresponding to “bicycle”, and a combination corresponding to “vehicle” and a region corresponding to “person”.

  According to such a configuration, there is a high possibility that such a combination is an accident involving the vehicle. Therefore, by specifying such a combination, the accident can be detected with higher accuracy. .

As a preferred configuration, the center transfers the accident image received from the first vehicle to a target engine corresponding to the accident.
According to such a configuration, it is expected that, for example, if it is a fire-fighting organization, it is possible to quickly and appropriately rescue and specify a transport destination by transferring such an accident image. It is expected that injuries will be grasped and appropriate treatment will be possible. For example, in the case of a police agency, quick safety and the like will be expected.

The block diagram which shows the schematic structure about 1st Embodiment of an accident image acquisition system. The sequence diagram which shows the outline of the flow which acquires the accident image in the accident image acquisition system. The flowchart which shows the flow of the process which recognizes the accident image in the accident image acquisition system. The figure which shows the aspect which partitions the captured image in the accident image acquisition system into a some area | region. The figure which shows the aspect which recognizes the content of the accident image from the captured image in the accident image acquisition system. The block diagram which shows the schematic structure about 2nd Embodiment of an accident image acquisition system. The sequence diagram which shows the outline of the flow which acquires the accident image in the accident image acquisition system. The block diagram which shows the schematic structure about 3rd Embodiment of an accident image acquisition system. The sequence diagram which shows the outline of the flow which acquires the accident image in the accident image acquisition system.

(First embodiment)
1st Embodiment which actualized the accident image acquisition system is described according to FIGS.

The outline of the present embodiment will be described with reference to FIGS. 1 and 2.
The accident image acquisition system shown in FIG. 1 includes a vehicle 20 as a first vehicle capable of capturing an accident image P1 of an accident vehicle 50 that has caused an accident with a relative object, and a second image captured by the vehicle 20. And a center 10 to which an accident image P1 including an accident vehicle 50 as a vehicle is transmitted. Further, the center 10 transmits the acquired accident image P1 to the target engine 30. The target engine 30 is an engine that provides rescue, medical care, and safety assurance with respect to an accident, and is an engine corresponding to an accident related to the accident vehicle 50, such as a fire engine, a medical institution, and a police organization. The vehicle 20 and the accident vehicle 50 are vehicles that travel on the road, and are cars such as passenger cars, taxis, buses, and trucks.

  As shown in FIG. 2, the vehicle 20 can image a traveling environment such as the front of the vehicle (step S <b> 10). Therefore, when an accident occurs between the accident vehicle 50 and the relative object in front of the traveling direction (step S11), the vehicle 20 captures an image including the accident vehicle 50 (step S12). Further, the vehicle 20 specifies the accident image P1 including the accident vehicle 50 and the relative object by performing image recognition processing on the captured image, and recognizes the occurrence of the accident by this specification (step S13). When the vehicle 20 specifies the accident image P1, the vehicle 20 transmits (uploads) the accident image P1 to the center 10 in a form including the position information and time information (step S14). Then, the center 10 transfers the accident image P1 and position information received from the vehicle 20 to an appropriate target engine 30 (step S15). The target engine 30 is provided with the accident image P1 and position information sent from the vehicle 20 different from the accident vehicle 50 via the center 10, so that there is no report from the accident vehicle 50 or its driver. Based on the provided information, rescue, medical care, and safety can be appropriately performed for the accident (step S16).

Next, the configuration of the accident image acquisition system will be described in detail.
As shown in FIG. 1, the target engine 30 includes a communication unit 31 that can communicate information with the center 10, an image acquisition unit 33 that acquires an accident image P <b> 1 and the like from the center 10 via the communication unit 31, 10 and a storage unit 32 that stores the accident image P1 acquired from the computer 10. The target engine 30 includes an image management unit 34 that manages the accident image P1 stored in the storage unit 32, and an image display unit 35 that displays the accident image P1 stored in the storage unit 32.

  In the target engine 30, the accident image P1 received from the center 10 is displayed on the image display unit 35. The accident image P1 is expected to enable quick and appropriate rescue and identification of an appropriate transport destination if the target engine 30 is a fire engine, and if the medical institution can quickly grasp the injury status, It is expected that appropriate treatment will be possible, and if it is a police agency, it is expected to be able to ensure quick safety.

  The center 10 acquired from the vehicle 20, the communication unit 11 that can communicate information with the vehicle 20 and the target engine 30, the image acquisition unit 13 that acquires the accident image P1 and the like from the vehicle 20 via the communication unit 11, and the vehicle 20. And a storage unit 12 for storing the accident image P1. The center 10 is stored in the storage unit 12 via the position specifying unit 14 that specifies the position where the accident occurred, the image management unit 15 that manages the accident image P1 stored in the storage unit 12, and the image management unit 15. And an image distribution unit 16 for providing (transferring) the accident image P1 to the target engine 30.

  The image acquisition unit 13 receives the accident image P <b> 1 from the vehicle 20 and stores it in the storage unit 12 via the image management unit 15. The center 10 recognizes the occurrence of the accident by receiving the accident image P1 from the vehicle 20. Further, the image acquisition unit 13 may recognize the received accident image P1, confirm that it is an accident, or grasp the situation of the accident from the image.

  The position specifying unit 14 acquires an accident point that is an accident occurrence point from the position information associated with the accident image P1. The position specifying unit 14 may specify the accident point as longitude / latitude, may be specified by a road link, or may be specified in association with road map information.

The image management unit 15 provides (transfers) to the target engine 30 via the image distribution unit 16 and the accident image P1. That is, the accident image P1 is automatically provided to the target engine 30.
Further, the image management unit 15 stores the accident image P1 in the storage unit 12 in such a manner that the search by time and position is possible. As the time, the imaged time given to the accident image P1 can be used. The image management unit 15 may cause the storage unit 12 to store the situation recognized by the image acquisition unit 13 by recognizing the accident image P1 in association with the accident image P1.

  Furthermore, the image management unit 15 can search the accident image P1 stored in the storage unit 12 based on conditions such as time and position. That is, when the time and position are designated, the image management unit 15 searches for the accident image P1 corresponding to the designated time and position, and outputs the accident image P1 as a search result.

  The image distribution unit 16 distributes the received accident image P1 to the target engine 30 corresponding to the accident. The image distribution unit 16 may select the target engine 30 corresponding to the accident indicated by the accident image P1 based on position information, time information, and the like. That is, the image distribution unit 16 may select an appropriate engine from a plurality of target engines 30 set in advance according to the position and time of the accident.

Incidentally, in the center 10, the operator may confirm the accident image P1, and the accident image P1 may be distributed to a predetermined target engine 30 in accordance with the operator's instruction.
The accident vehicle 50 is a vehicle involved in an accident with a counterpart that is the other party of the accident. The accident vehicle 50 is in a state where it can be imaged by the imaging unit 24 of the vehicle 20. Examples of the accident between the accident vehicle 50 and the relative object include a contact accident, a rear-end collision, and a collision accident.

  The vehicle 20 includes a communication unit 21 capable of communicating information with the center 10 and a position specifying unit 23 that specifies the current position of the vehicle 20 as position information. In addition, the vehicle 20 includes an imaging unit 24 that images the surrounding environment of the vehicle 20, a storage unit 22 that stores a captured image including the accident image P <b> 1 captured by the imaging unit 24, and an accident image stored in the storage unit 22. And an image management unit 26 that manages captured images including P1. Furthermore, the vehicle 20 includes an image recognition unit 25 that performs image recognition processing on the captured image captured by the imaging unit 24.

Note that the accident image acquisition system of the present embodiment includes at least the image recognition unit 25 of the vehicle 20, the image management unit 26, and the image acquisition unit 13 of the center 10.
The position specifying unit 23 specifies the current position of the vehicle 20 using a GPS (global positioning system) device or the like.

  The imaging unit 24 is a camera such as a CCD camera that can image a traveling environment in front of the vehicle 20. The imaging unit 24 automatically images the traveling environment periodically or based on predetermined conditions. The imaging unit 24 associates position information and time information with the captured image.

  Here, the relative positional relationship between the vehicle 20 and the accident vehicle 50 will be described. The relative positional relationship between the vehicle 20 and the accident vehicle 50 is preferably an arrangement in which an accident between the accident vehicle 50 and a relative object can be determined from a captured image of the imaging unit 24 of the vehicle 20. For example, it is preferable that the imaging range of the imaging unit 24 includes the side surface of the accident vehicle 50 in contact with a relative object. Such a relative positional relationship between the vehicle 20 and the accident vehicle 50 is likely to occur at a place where roads intersect such as an intersection or a side road entrance / exit. If it is an accident at an intersection, the vehicle 20 can see the face where the accident of the accident vehicle 50 occurred from any road. It is also preferable that the accident vehicle 50 and the relative object are included in the imaging range of the imaging unit 24 of the vehicle 20.

  The image management unit 26 receives a captured image that has undergone image recognition processing from the image recognition unit 25. The image management unit 26 transmits the accident image P1 to the center 10 when the input image is the accident image P1. Therefore, the image management unit 26 transmits the image sent as the accident image P1 from the image recognition unit 25 to the center 10 through the communication unit 21 together with the position information and the time information. That is, the image management unit 26 automatically transmits the accident image P1 to the center 10.

  In addition, the image management unit 26 may cause the storage unit 22 to store a captured image including the accident image P <b> 1 input from the image recognition unit 25. Further, the image management unit 26 may store the captured image that has not been image-recognized from the imaging unit 24 in the storage unit 22.

  The image recognition unit 25 receives the captured image captured by the imaging unit 24, and performs image recognition of the input captured image to extract information that is significant for driving support. In the present embodiment, the image recognition unit 25 performs a process for recognizing an accident occurring in the traveling environment. In the image recognition unit 25, the image recognition process includes an area division process and a content classification process.

  The recognition processing in the image recognition unit 25 will be described in detail with reference to FIGS. The recognition process shown in FIG. 3 is started each time a captured image captured by the imaging unit 24 is input.

  When the captured image is input, the image recognizing unit 25 divides the input captured image into regions for each segmentation including a group of pixels having the same properties (step S121 in FIG. 3). As a result, the captured image is divided into a plurality of regions. That is, each region is formed by grouping pixels having similar properties among the pixels of the captured image.

  FIG. 4 shows an example of the image P divided into regions. In the image P, eleven areas D10 to D20 are formed by area division. Each of the regions D10 to D20 is composed of a pixel having a small difference in properties, that is, a collection of similar pixels, while each of the regions D10 to D20 includes a pixel having a large difference in properties, that is, a collection of so-called dissimilar pixels. Become. In the present embodiment, the contents of the formed regions D10 to D20 are not specified only by the region division processing.

  Subsequently, the image recognition unit 25 identifies the contents of the regions D10 to D20 (step S122 in FIG. 3). That is, the image recognition unit 25 associates one appropriate content with each of the regions D10 to D20. The contents include “vehicle”, “motorcycle”, “person”, “bicycle”, “mountain”, “sky”, “ground”, “building”, and the like. In this embodiment, the “ground” includes a road surface, the “vehicle” has a vehicle body composed of a metal surface or a glass surface having a predetermined area, and the “motorcycle” is more than the “vehicle”. And a vehicle body composed of a metal member or a resin member having a narrow area.

  FIG. 5 shows the contents of the specified areas D10 to D20. That is, each of the areas D10, D13, D14, and D16 is specified as “ground”, the area D11 is specified as “vehicle”, and the area D12 is specified as “motorcycle”. Further, the contents of the area D15 are specified as “person”, the areas D17 and D18 as “buildings”, the area D19 as “mountains”, and the area D20 as “sky”.

  By the way, in this embodiment, the content of each area | region D10-D20 is specified using classifiers, such as a support vector machine (SVM). Therefore, a content classification method will be described by taking SVM as an example. In addition, although SVM is demonstrated easily here, it is applicable also to the classification method using SVM.

  In SVM, for each of a plurality of contents, input data (teacher data) Xi (vector) corresponding to the contents and contents yi (scalar) to be classified are determined (i = 1 to m: m corresponds to the number of classifications). Natural number). The input data (teacher data) Xi is a vector composed of a plurality of elements (xi1, xi2, xi3,... Xin: elements are scalars and n is the number of elements) including information related to the contents. At this time, when the function f (X) composed of the input data (teacher data) Xi and the parameter W is defined as in the formula (1), the formula (1) is given the input data (teacher data) Xi. A parameter W for calculating the corresponding content yi is determined. The index “T” of the parameter W in Expression (1) indicates “transposition”.

f (Xi) = W ^T Xi (1)
According to equation (1), when unknown input data Xj other than the teacher data Xi is given to the right side, a value corresponding to the given input data Xj is obtained from the function f (Xj). Then, by comparing the value of the function f (Xj) with each content yi and estimating the content yi corresponding to the value of the function f (Xj), the content of the input data Xj is determined from the value of the function f (Xj). To be estimated. Therefore, according to the classification method of SVM, by using the parameter W, the unknown input data Xj is converted into “vehicle”, “motorcycle”, “person”, “mountain”, “sky”, “ground” and “building”. It can be estimated to any one of the contents.

  Next, the image recognition unit 25 performs accident recognition (step S123 in FIG. 3). When it is determined that the accident is not recognized (NO in step S123 in FIG. 3), the image recognition unit 25 ends the image recognition process. This image recognition process is started when a new image is captured. On the other hand, if it is determined that an accident has been recognized (YES in step S123 in FIG. 3), the image recognition unit 25 uploads the position information and the accident image P1 to the center 10 (step S124 in FIG. 3), and then the image. The recognition process ends.

The accident recognition performed in step S123 will be described in detail.
First, the image recognition unit 25 detects the distance from the vehicle 20 to each of the regions D10 to D20. Each distance may be detected by the imaging unit 24, may be detected by an on-vehicle radar (not shown), or may be detected from a captured image captured by the imaging unit 24. For example, when detecting from the captured image, the distance from the vehicle 20 can be detected based on the fact that the ground below the image P is a position close to the vehicle 20 and the center ground is a position far from the vehicle. it can. That is, the area corresponding to the contents such as “vehicle”, “motorcycle”, “person”, “bicycle”, “building”, etc. in contact with the ground is the distance from the lower side of the image P to the position in contact with the ground. Based on the above, the distance from the vehicle 20 can be calculated. Referring to FIG. 5, for example, the distance from the vehicle 20 to the region D11 is calculated based on the length L11 to the portion in contact with the ground in the region D11 and the length L21 to the portion facing the ground. Is done. Further, the distance from the vehicle 20 to the region D12 is calculated based on the length L12 in contact with the ground in the region D12. Here, since the length L12 of the area D12 is included between the length L11 and the length L21 of the area D11, the distance between the area D11 and the area D12 overlaps, that is, partly has the same distance. .

  When the distance of each of the areas D10 to D20 from the vehicle 20 is detected, the image recognition unit 25 identifies a combination of areas that are at the same distance and at least one of the areas is “vehicle”. This is because an accident involving the “vehicle” is recognized, and the “vehicle” and the relative object are considered to have substantially the same distance from the vehicle 20. For example, in FIG. 5, the region D <b> 11 is specified as the “vehicle” region. Then, the region D12 is specified as a region at the same distance as the region D11. Thus, the combination of the region D11 and the region D12 is specified as a combination of regions in which at least one of the adjacent regions is “vehicle”.

  Subsequently, it is determined whether or not an accident has occurred between the “vehicle” and the relative object for the specified combination of regions. In the present embodiment, whether or not an accident has occurred is determined based on the degree of proximity between the “vehicle” area and the relative object area. At this time, as a combination of the areas to be judged as accidents, it corresponds to the combination of the area corresponding to the vehicle (hereinafter referred to as “vehicle” area) and the “vehicle” area, and corresponds to the “vehicle” area and the motorcycle. And a combination with a region to be performed (hereinafter referred to as a “bike” region). Further, as a combination of adjacent areas, a combination of a “vehicle” area and an area corresponding to a bicycle (hereinafter referred to as “bicycle” area), an area corresponding to a “vehicle” area and a person (hereinafter referred to as “person”). And the combination of “region”.

  Then, the occurrence of an accident is determined based on the degree of proximity between the two areas for these specified combinations. For example, if the distance between the “vehicle” area and the “vehicle” area, the “bike” area, the “bicycle” area, and the “person” area, which are relative object areas, is high, It is recognized that an accident has occurred when the predetermined distance is determined. It should be noted that the predetermined distance determined to have a high degree of proximity includes a state where there is no distance, that is, a state of contact (a high degree of proximity), or an unusual proximity, for example, a few centimeters or a few tens of centimeters Can include metric status. Such a distance may be provided theoretically, experimentally, or empirically, or may be provided statistically from accident statistics or the like. Also, even if the distance is a little, the height of the “bike” area, “bicycle” area, and “person” area is low, that is, the motorcycle, the bicycle, or the person is lying down. You may recognize that an accident has occurred. Furthermore, in the case of the “vehicle” area and the “vehicle” area, even if there is some distance, it is recognized that an accident has occurred when the side that normally does not face on the road and the front and rear faces each other. May be.

  For example, in FIG. 5, the area D11 and the area D12 are the “vehicle” area and the “motorcycle” area, respectively, and are subject to the occurrence of an accident, and the distance D is only a distance DS between the area D11 and the area D12. Therefore, it is recognized that the degree of proximity is high. Thereby, it is recognized that an accident has occurred. In FIG. 5, the distance DS is about the thickness of the motorcycle tire. The closer the distance DS is to “zero”, the higher the proximity degree, and the longer the distance, the lower the proximity degree. Note that the distance DS (degree of proximity) at the time of being recognized as an accident can be selected from “zero” from a distance that cannot be stopped based on the relative speed.

When the image recognition unit 25 recognizes an accident from the captured image by the image recognition process, the image recognition unit 25 outputs the image as an accident image P1, and the image management unit 26 recognizes the occurrence of the accident and the accident image P1.
With reference to FIG. 2, the operation of the accident image acquisition system of the present embodiment will be described.

  The vehicle 20 images the traveling environment ahead of the vehicle 20 at predetermined intervals in response to the start of traveling (step S10 in FIG. 2). This imaging is automatically repeated from the start of travel to the end of travel. The imaging interval may range from several tens of frames per second to one frame per few seconds. The vehicle 20 performs image processing on the captured image as necessary to acquire information that is significant for driving support and recognizes the accident, but the captured image is not accident-recognized and is only stored in the storage unit 22. May be included.

  When an accident involving the accident vehicle 50 different from the own vehicle occurs in front of the traveling direction of the vehicle 20 (step S11 in FIG. 2), the vehicle 20 takes an image including the accident vehicle 50 (FIG. 2). Step S12). The vehicle 20 recognizes the captured image including the accident vehicle 50 to identify the accident image P1 in which the accident involving the accident vehicle 50 is captured, that is, recognizes the occurrence of the accident (step of FIG. 2). S13).

  When the accident image P1 is identified from the captured image, the vehicle 20 automatically transmits (uploads) the identified accident image P1 to the center 10 together with the corresponding position information and time information (step S14 in FIG. 2). That is, in the vehicle 20, when the image recognition unit 25 recognizes the accident image P1, the image management unit 26 automatically transmits the accident image P1 including the position information and time information to the center 10 via the communication unit 21. Done. Note that when automatically transmitting in this way, the image management unit 26 may or may not store the accident image P1 in the storage unit 22.

  The center 10 that has received (acquired) the accident image P1 automatically transfers the received (acquired) accident image P1 to the target engine 30 (step S15 in FIG. 2). That is, the accident is quickly notified to the target engine 30 in a manner in which the processing load on the center 10 is small. The center 10 may hold and manage the acquired accident image P1 at the center 10, or may perform transfer or the like after performing predetermined processing. In this case, if the accident is determined based on a plurality of accident images, the accuracy of the accident notification can be increased.

  Since the target organization 30 can grasp the detailed situation of the accident that is difficult to grasp only by voice notification based on the accident image P1 received from the center 10, it is possible to rescue, medical, safety Securement can be performed appropriately (step S16 in FIG. 2). Thereby, the target engine 30 can grasp the accident even when there is no report from the accident party such as the accident vehicle 50 or its driver. Moreover, the situation about the accident can be grasped more suitably by the accident image P1.

  In the present embodiment, when the vehicle 20 recognizes the accident, the vehicle 20 automatically transmits the accident image P1 to the center 10. Therefore, even if the accident vehicle 50 cannot make the accident report, the accident report or rescue is performed. Images for treatment and ensuring safety will be provided to the center. When the accident vehicle 50 is unable to report an accident, the accident is detected when the device for notifying the accident is not provided or the impact of the accident is weak while the device for notifying the accident is provided. Such as when it was not possible, or when a device for reporting an accident broke down. In addition, for example, the impact of an accident may not be so strong due to contact with a pedestrian or a bicycle. In such a case, there is a possibility that an accident detection based on the impact may not be performed. 10 will be provided.

According to the present embodiment, it is possible to provide an accident image acquisition system that can cause the center 10 to appropriately acquire the accident image P1 captured through the vehicle 20.
As described above, the accident image acquisition system according to the present embodiment has the effects listed below.

  (1) By transmitting the captured image recognized as the accident image P1 from the vehicle 20 to the center 10, the center 10 can acquire a significant captured image for grasping the situation of the accident involving the accident vehicle 50. It becomes like this.

  Further, even if the accident vehicle 50 cannot notify the center 10 of the situation of the accident, the accident image P1 captured by the vehicle 20 is appropriately acquired by the center 10 as long as the vehicle 20 exists in the vicinity of the accident vehicle 50. Become. Then, the center 10 can recognize an accident, report an accident, and hold an image significant for grasping the situation of the accident. That is, if the accident is recognized at the center 10, even if the accident vehicle 50 is unable to report the accident, an image for accident notification, rescue, treatment, and safety is provided to the center. become. As for the case where it is not possible to report an accident, it is not possible to detect an accident because it is not equipped with a device for reporting an accident or is equipped with a device for reporting an accident, but the impact of the accident is weak. Or when a device for reporting an accident breaks down.

  Furthermore, since the captured image recognized as the accident image P1 is acquired by the center 10, compared to a case where a large amount of non-accident images are included and a necessary image cannot be obtained unless screening is performed. The value of the image acquired by the center 10 is naturally increased. Moreover, since only the captured image recognized as the accident image P1 is transmitted, the communication amount can be suppressed.

  In addition, the accident image P1 can be appropriately recognized from the captured image by recognizing the accident image P1 based on the degree of proximity between the accident vehicle 50 and a relative object close to the accident vehicle 50. .

  (2) Since the accident image is automatically transmitted to the center 10 based on the determination of the accident recognition in the vehicle 20, an accident notification or the like is realized based on the recognition of the accident image P1 of the vehicle 20. .

  (3) By transmitting the accident image P1 from the vehicle 20 that is estimated to have captured the accident in the vicinity of the accident point, the center 10 can acquire and acquire an appropriate accident image. As a result, the center 10 can acquire the necessary accident image P <b> 1 while suppressing an increase in the amount of communication with the vehicle 20.

  (4) According to the process of creating an area (superpixel or segmentation) by dividing an image according to features, and the process of specifying that “vehicle” is included in the created combination of divided areas, The load on the image recognition process can be reduced. And the accident vehicle 50 which caused the accident out of this combination comes to be identified.

(5) Appropriate image classification can be performed by the support vector machine.
(6) Since the combination when recognizing the accident defined as described above is likely to be an accident involving "vehicle", the accident is detected with higher accuracy by specifying such a combination. Will be able to.

  (7) By transferring such accident images, for example, if it is a fire engine, it is expected that it will be possible to quickly and appropriately rescue and specify the transport destination. It is expected that appropriate treatment will be possible. For example, in the case of a police agency, prompt safety is expected.

(Second Embodiment)
A second embodiment of the accident image acquisition system will be described with reference to FIGS. 6 and 7.

  Prior to the transmission of the accident image P1 from the vehicle 20, the accident image acquisition system of the present embodiment notifies the center 10 of the location information of the accident point, and the center 10 selects the vehicle 20 and the other vehicle 40 selected based on the accident point. This is different from the accident image acquisition system of the first embodiment in that the accident images P1 and P2 are acquired from the configuration. However, since the configurations other than the above-described differences are the same, the same reference numerals are given to the similar configurations, and the detailed description thereof is omitted.

  The image acquisition unit 13 of the center 10 can further request the vehicle 20 to transmit the accident image P1, and can receive the accident image P1 from the vehicle 20 as a response to the request.

The position specifying unit 14 can further specify the accident point from the position information sent independently from the vehicle 20 instead of being associated with the accident image P1.
The center 10 further includes a vehicle selection unit 17 that requests the vehicle 20 or the like to transmit an accident image. When the accident information is input, the vehicle selection unit 17 selects the vehicle 20 or the other vehicle 40 that is traveling near the place where the accident occurs based on the position information of each vehicle that the center 10 grasps. The vehicle 20 and the other vehicle 40 are requested to transmit the accident image. The time of occurrence is also notified. Further, the center 10 acquires the accident image transmitted from the vehicle 20 or the other vehicle 40 that requested the transmission of the accident image. Note that by notifying the occurrence time, the possibility that an accident image to be acquired can be acquired is further increased. In addition, even if the elapsed time from the occurrence of the accident is long, if the occurrence time is notified, an appropriate accident image can be acquired, and the time required for searching the accident image in the vehicle that requested the accident image Can be shortened.

  The other vehicle 40 has the same configuration as the vehicle 20. That is, the communication unit 41, the storage unit 42, the position specifying unit 43, the imaging unit 44, the image recognition unit 45, and the image management unit 46 of the other vehicle 40 are respectively connected to the communication unit 21, the storage unit 22, and the position specifying unit 23 of the vehicle 20. Since it has the same configuration as the imaging unit 24, the image recognition unit 25, and the image management unit 26, a detailed description is omitted.

  The vehicle 20 and the other vehicle 40 appropriately notify the center 10 of position information that can specify the position. As described above, as a vehicle for notifying the center 10 of position information, there is a probe car for uploading (notifying) the road condition to the center 10.

  Further, the image management unit 26 can search and output the captured image stored in the storage unit 22 based on position information and time information. That is, a specific accident image P1 is retrieved from the storage unit 22 based on the position information and time information. Therefore, when receiving the “request to transmit the accident image” from the center 10, the image management unit 26 searches the storage unit 22 for the accident image P 1 that meets the conditions of the position information and time information included in the request, The searched accident image P1 is transmitted to the center 10.

  The vehicle 20 and the other vehicle 40 transmit the recognized accident images P <b> 1 and P <b> 2 to the center 10 in response to the accident image transmission request. Further, the vehicle 20 and the other vehicle 40 transmit the identified accident images P1 and P2 to the center 10 by performing a recognition process of the captured image stored in response to the accident image transmission request.

  That is, in the vehicle 20 and the other vehicle 40, the use value as an image significant for grasping the situation regarding the accident is naturally increased by specifying the accident images P1 and P2. Further, by specifying the accident images P1 and P2, it is possible to reduce the time required for searching the accident images P1 and P2 as compared with the case of searching from a captured image in which no accident image is specified.

The operation of the accident image acquisition system will be described with reference to FIG.
When an accident occurs (step S11 in FIG. 7), the vehicle 20 photographing the front of the vehicle (step S11 in FIG. 7) takes an accident image P1 including the accident vehicle 50 (step S12 in FIG. 7). An accident is recognized based on the image recognition of the accident image P1 (step S13 in FIG. 7). When the vehicle 20 recognizes the accident, the vehicle 20 notifies (uploads) the location information of the accident point to the center 10 (step S20 in FIG. 7). On the other hand, the other vehicle 40 appropriately captures an image during traveling (step S40 in FIG. 7), and notifies (uploads) the position information of the own vehicle to the center 10 as appropriate (step S21 in FIG. 7).

  When the location information of the accident location is notified, the center 10 selects the vehicle 20 and the other vehicle 40 that are close to the accident location among the vehicles that grasp the travel location as the vehicle for transmitting the accident image, and the selected vehicle 20 And other vehicles 40 are requested to transmit accident images (steps S22 and S23 in FIG. 7).

  The vehicle 20 and the other vehicle 40 that have received the transmission request of the accident image from the center 10 respectively transmit (upload) the accident images P1 and P2 that are retrieved from the captured images as satisfying the conditions to the center 10 ( Steps S25 and S26 in FIG.

  Then, the center 10 provides the received accident image to the target engine 30 (step S15 in FIG. 7). Examples of provision include rapid transfer of received accident images, provision after predetermined processing, and the like. Therefore, the target engine 30 appropriately performs rescue, medical care, and safety assurance based on the provided accident image (step S16 in FIG. 7).

  In the present embodiment, the center 10 can recognize the accident of the accident vehicle 50 by obtaining the accident images P <b> 1 and P <b> 2 from the vehicle 20 or the other vehicle 40 different from the accident vehicle 50. The center 10 narrows down and acquires the accident images P1 and P2, thereby reducing the load on the acquisition. Since the vehicle 20 and the other vehicle 40 transmit the accident images P1 and P2 recognized from the captured images to the center 10, a large amount of non-accident images are included and a necessary image is obtained unless screening is performed. The processing load is reduced as compared to the case where it is not possible. Further, since only the accident image P1 is transmitted from the vehicle 20 or the other vehicle 40 to the center 10, the amount of communication can be suppressed.

As described above, the accident image acquisition system according to this embodiment has the effects listed below in addition to the effects (1) to (7) described in the first embodiment.
(8) By transmitting the position information, the vehicle 20 enables a quicker and more reliable notification of an accident than the transmission of a captured image. In addition, immediately after the accident is recognized, it is possible to suppress a situation in which captured images are transmitted all at once from a plurality of vehicles and the amount of communication increases rapidly.

(9) By transmitting the accident image P <b> 1 in response to a request from the center 10, it is possible to reduce the processing load related to the transmission of the image of the vehicle 20 or the other vehicle 40.
(Third embodiment)
A third embodiment embodying the accident image acquisition system will be described with reference to FIGS.

  In the second embodiment, the accident image acquisition system of the present embodiment is configured such that the accident vehicle 50 notifies the center 10 of the position information of the accident point, and the center 10 acquires the accident image from the vehicle 20 close to the accident point. It differs from the accident image acquisition system of the form. However, since the configurations other than the above-described differences are substantially the same, the same reference numerals are given to the substantially similar configurations, and the detailed description thereof is omitted.

  The accident vehicle 50 has the same configuration as the vehicle 20. That is, the communication unit 51, the storage unit 52, the position specifying unit 53, the imaging unit 54, the image recognition unit 55, and the image management unit 56 of the accident vehicle 50 are respectively connected to the communication unit 21, the storage unit 22, and the position specifying unit 23 of the vehicle 20. Since it has the same configuration as the imaging unit 24, the image recognition unit 25, and the image management unit 26, a detailed description is omitted.

  The accident vehicle 50 detects the occurrence of an accident in the own vehicle through an accident detection device including an impact sensor (not shown) and notifies the center 10 of the position information of the accident point. Thereby, the center 10 specifies an accident point. The accident vehicle 50 may transmit an accident image P3 that is an accident image captured by the own vehicle. Since the accident image P3 is a captured image from inside the vehicle, it is difficult to grasp the situation of the portion near the vehicle body where the vehicle can be recognized from the outside.

The center 10 selects the vehicle 20 traveling near the accident point by the vehicle selection unit 17 and requests the selected vehicle 20 to transmit the accident image.
The vehicle 20 transmits the recognized accident image P1 to the center 10 in response to the accident image transmission request.

The operation of the accident image acquisition system will be described with reference to FIG.
The vehicle 20 is photographing the front (step S10 in FIG. 9). When an accident occurs (step S11 in FIG. 9), the accident vehicle 50 detects the occurrence of an accident in its own vehicle and transmits (uploads) the location information of the accident point to the center 10 (step S30 in FIG. 9).

  The vehicle 20 captures the accident image P1 in the vicinity of the accident vehicle 50 (step S12 in FIG. 9), and recognizes that an accident has occurred based on image recognition of the captured image (step S13 in FIG. 9). Further, the vehicle 20 appropriately notifies (uploads) the position information of the own vehicle to the center 10 (step S20 in FIG. 9).

  When the location information of the accident point is notified, the center 10 selects the vehicle 20 as an image uploading vehicle from among the vehicles grasped by the center 10 and transmits the accident image to the selected vehicle 20. A request is made (step S22 in FIG. 9). The vehicle 20 transmits (uploads) the accident image P1 to the center 10 in response to the request for uploading the accident image from the center 10 (step S25 in FIG. 9).

  Then, the center 10 provides the received accident image to the target engine 30 (step S15 in FIG. 9). The target engine 30 appropriately performs rescue, medical care, and safety assurance based on the provided accident image (step S16 in FIG. 9).

As described above, the accident image acquisition system according to the present embodiment has the effects listed below in addition to the effects (1) to (9) described in the first and second embodiments.
(10) Even if the accident vehicle 50 reports an accident, the accident image P1 obtained by imaging the accident vehicle 50 from the outside can be acquired, so that the situation of the accident can be grasped in more detail.

(Other embodiments)
In addition, each said embodiment can also be implemented with the following aspects.
In the second and third embodiments, the case where the center 10 requests transmission of an accident image to the vehicle 20 or the other vehicle 40 whose position is known has been illustrated. However, the center is not limited to this, and the center notifies the surrounding vehicles of the accident occurrence location and the time at which the accident occurred, and requests that an accident image be sent if there is a vehicle that has passed through the accident occurrence location after the accident occurrence time. Also good. Thereby, even if there is a vehicle whose position is not grasped, an accident image can be acquired. Thereby, the application range of the accident image acquisition system can be expanded.

  In the second and third embodiments, the case where the occurrence time is notified when the transmission of the accident image is requested is illustrated. However, the present invention is not limited to this, and the generation time may not be notified. Since the location information of the accident site is notified and the transmission of the accident image is requested, if the elapsed time from the occurrence of the accident is short, an appropriate accident image is highly likely to be acquired. Thereby, the application range of the accident image acquisition system can be expanded.

  In the second and third embodiments, the case where the vehicle 20 transmits the recognized accident image P1 to the center 10 is illustrated. However, the present invention is not limited to this, and in response to receiving the accident image transmission request from the center, the vehicle performs image recognition on the captured image obtained from the imaging unit and the captured image stored in the storage unit, and displays the accident image. It may be specified. Thereby, the application range of the accident image acquisition system can be expanded.

  In each of the above embodiments, the case where the accident image P1 is provided from the center 10 to the target engine 30 is illustrated. However, the present invention is not limited to this, and the center may be the target engine. Thereby, the activity based on the received accident image can be performed more quickly.

  In each of the above-described embodiments, the center 10 is exemplified for the case where the image acquisition unit 13, the position specifying unit 14, the image management unit 15, the image distribution unit 16, and the like are provided separately. However, as long as these functions of the image acquisition unit, the position specifying unit, the image management unit, and the image distribution unit are performed, even if these are processed by one arithmetic device, they are processed by a plurality of arithmetic devices. It may be. The same applies to the target engine 30. As a result, the degree of freedom in designing the accident image acquisition system can be improved.

  In each of the above-described embodiments, the vehicle 20 is illustrated with respect to the case where the position specifying unit 23, the image recognition unit 25, the image management unit 26, and the like are provided separately. However, as long as these functions of the position specifying unit, the image recognition unit, and the image management unit are performed, even if these are processed by one arithmetic device, they are processed by a plurality of arithmetic devices. May be. The same applies to the other vehicle 40 or the accident vehicle 50. As a result, the degree of freedom in designing the accident image acquisition system can be improved.

  In each of the above embodiments, an apparatus in which the position specifying unit 23, the image recognition unit 25, the image management unit 26, and the like are executed in the vehicle 20 is not particularly limited. That is, each of these units may be executed by an on-board electronic control unit (ECU) or a navigation system, or may be executed by a portable information processing device that is brought into the vehicle and used. Examples of portable information processing apparatuses include mobile phones, smartphones, and portable navigation systems.

The imaging unit 24 may also be a camera provided in a vehicle or a camera provided in a portable information processing apparatus or the like.
The same applies to the other vehicle 40 or the accident vehicle 50 with respect to the apparatus and the imaging unit in which the position specifying unit, the image recognition unit, the image management unit, and the like described above are executed.

As a result, the degree of freedom in designing the accident image acquisition system can be improved.
In each of the above embodiments, the case where the imaging unit 24 images the traveling environment in front of the vehicle 20 is illustrated. However, the present invention is not limited to this, and the imaging unit may include a direction other than the front, such as the rear, right direction, and left direction of the vehicle. If an accident image can be captured regardless of the direction of image capture, the accident image can be transmitted from the vehicle to the center. Thereby, the application range of the accident image acquisition system can be expanded.

  In each of the above embodiments, the case where region division is performed by segmentation and content classification is performed by SVM is illustrated. However, the present invention is not limited to this, and if an accident image can be specified from a captured image, region division or content classification may be performed by other methods, or image recognition processing may be performed by pattern matching or the like. As a result, the degree of freedom in designing the accident image acquisition system can be improved.

  In the above embodiments, the contents classified are “vehicle”, “motorcycle”, “person”, “bicycle”, “mountain”, “sky”, “ground”, “building”, and the like. The case was illustrated. However, not limited to this, the contents to be classified include road equipment such as “curbstone”, “side wall”, “signpost” and “post”, “electric pole”, “river”, “tabata”, etc. Also good. The situation can be recognized in detail by increasing the contents to be classified. Further, if the contents to be classified are narrowed down, the load required for the recognition process can be reduced. Thereby, the application range of the accident image acquisition system can be expanded.

  In each of the above embodiments, the case where the vehicle is an “automobile” (excluding a motorcycle) is exemplified. However, the present invention is not limited to this, and the vehicle may include a “motorcycle” and a “bicycle”. If “motorcycle” or “bicycle” is included, an accident between “motorcycle” and “motorcycle”, “bicycle” or “person”, “bicycle” and “bicycle” or “person” You may make it detect the accident between and. Thereby, the application range of the accident image acquisition system can be expanded.

  DESCRIPTION OF SYMBOLS 10 ... Center, 11, 21, 31, 41, 51 ... Communication part, 12, 22, 32, 42, 52 ... Storage part, 13, 33 ... Image acquisition part, 14 ... Position specification part, 15 ... Image management part, DESCRIPTION OF SYMBOLS 16 ... Image delivery part, 17 ... Vehicle selection part, 20 ... Vehicle, 23, 43, 53 ... Position specification part, 24, 44, 54 ... Imaging part, 25, 45, 55 ... Image recognition part, 26, 46, 56 ... Image management part, 30 ... Target organization, 34 ... Image management part, 35 ... Image display part, 40 ... Other vehicle, 50 ... Accident car, P1, P2, P3 ... Accident image, D10-D20 ... area.

Claims (8)

An accident image acquisition system comprising: a first vehicle that transmits a captured image obtained by imaging the periphery of the vehicle; and a center that acquires a captured image transmitted from the first vehicle,
The first vehicle includes a recognition unit that recognizes an accident image involving a second vehicle different from the host vehicle from the captured image, and the captured image recognized as the accident image by the recognition unit is the center. To send to
When the recognizing unit determines that the second vehicle and a relative object close to the second vehicle are present in the image through image processing of the captured image, the degree of proximity between the second vehicle and the relative object An accident image acquisition system for recognizing whether or not the captured image is an accident image based on the above. The accident image acquisition system according to claim 1, wherein the first vehicle automatically transmits the recognized accident image to the center in response to recognizing the accident image from the captured image. The accident image acquisition system according to claim 1, wherein the first vehicle transmits position information corresponding to the accident image to the center prior to transmission of the recognized accident image. The center requests the transmission of the accident image to the first vehicle estimated to have captured the accident image based on the received position information,
The accident image acquisition system according to claim 3, wherein the first vehicle requested to transmit transmits an accident image in response to a transmission request from the center. The recognizing unit divides the captured image into a plurality of regions based on the characteristics of the pixels, and identifies a combination in which at least one of the adjacent regions is “vehicle” from the divided regions. The accident image acquisition system according to any one of claims 1 to 4, wherein an accident image is recognized based on a proximity degree between two regions corresponding to the combination. The accident according to claim 5, wherein the recognition unit performs image classification of each region by image processing using a support vector machine, and identifies a combination in which at least one of adjacent regions is "vehicle" based on the image classification. Image acquisition system. The combination corresponds to a combination of an area corresponding to “vehicle” and an area corresponding to “vehicle”, a combination of an area corresponding to “vehicle” and an area corresponding to “motorcycle”, and “vehicle”. The accident according to claim 5 or 6, including at least one combination of a region and a region corresponding to "bicycle", and a combination corresponding to "vehicle" and a region corresponding to "person". Image acquisition system. The accident image acquisition system according to any one of claims 1 to 7, wherein the center transfers the accident image received from the first vehicle to a target engine corresponding to the accident.

Images