JP6573775B2 - Intersection monitoring system - Google Patents

Description

  The present invention relates to a technique for monitoring the movement of a moving object passing through an intersection.

  There is a technique for monitoring the movement of a moving object passing through an intersection. For example, Patent Document 1 describes a method for determining whether or not the movement of a vehicle at an intersection is an abnormal movement or a movement in violation of a rule by using a neuro-network by using an integrated block method. In the method described in Patent Document 1, a neuro-network that learns the relationship between the abnormal movement of a vehicle in an intersection or a phenomenon of rule violation and the total amount of movement of each block is obtained from each actual movement of the vehicle. The total motion amount of the blocks is compared with the learning content, and it is determined whether or not the actual vehicle motion is an abnormal motion or a motion in violation of the rules.

JP 2006-285399 A

  The movement of a moving object passing through an intersection is not uniform. Therefore, in order to determine whether or not the movement of the moving body is normal, it is necessary to determine whether or not many possible movements are normal. When this determination is performed by a human, it requires a great deal of labor. Moreover, even if this determination is performed by the apparatus, human beings must perform the work of confirming the validity of the result. The work for confirming the validity of the result of the determination performed by the apparatus also has a great effort.

  In view of the above circumstances, an object of the present invention is to reduce the labor required to monitor the movement of a moving body that passes an intersection.

In order to solve the above-described problem, the present invention provides a trajectory data of a moving body passing through the intersection indicated by a moving image captured by an imaging device installed at an intersection, the type of the intersection, and the current state of the traffic signal at the intersection. an acquiring unit that acquires the determination reference data indicative of the reference of the moving object behavior determination in accordance with the combination of steps before Kigen示階ladder at the time of the type and the passage of示階ladder, the trace data in light of the above determination reference data An intersection monitoring system comprising a determination unit that determines whether the movement indicated by is normal is provided as a first aspect.

According to the intersection monitoring system of the first aspect, the movement of the moving body that passes through the intersection is determined according to the type of the intersection, the type of the displayed floor of the traffic light at the intersection, and the step of the displayed floor at the time of passing. It is judged against light. Therefore, compared with the case where the criterion is not used, the effort required to monitor the movement of the moving body passing through the intersection is reduced.

  In the intersection monitoring system according to the first aspect, the acquisition means acquires timing data indicating the timing at which the moving body passes the intersection in the step of the current floor of the traffic light, and the acquisition means includes the timing A configuration in which the determination reference data according to data is acquired may be employed as the second aspect.

  According to the intersection monitoring system of the second aspect, the movement of the moving body passing through the intersection is determined in light of the determination criterion corresponding to the passing timing of the moving body in the present step. Therefore, compared with the case where the determination criterion according to the passing timing of the moving body in the present step is not used, the labor required for the determination criterion creation work and the determination result confirmation work is reduced.

  In the intersection monitoring system according to the first or second aspect, the acquisition unit acquires the present state data indicating the present state of the traffic signal at the intersection at the time when the moving body passes the intersection, and the acquisition unit includes: A configuration in which the determination reference data according to the display state data is acquired may be employed as a third aspect.

  According to the intersection monitoring system of the third aspect, the determination reference data is acquired according to the present state data acquired from, for example, a signal control device that controls display switching of traffic lights. Therefore, it is not necessary for the intersection monitoring system to manage the display state independently.

  In the intersection monitoring system according to any one of the first to third aspects, the acquisition unit acquires a plurality of trajectory data and generates the determination reference data based on the plurality of trajectory data acquired by the acquisition unit. A configuration including means may be adopted as the fourth aspect.

  According to the intersection monitoring system of the fourth aspect, the determination reference data is generated based on the movement of the moving object indicated by the plurality of trajectory data. Therefore, the time and effort for the user to create the determination reference data is reduced.

  In the intersection monitoring system according to the fourth aspect, the acquisition unit acquires determination result data indicating whether or not the movement of the moving body indicated by each of the plurality of trajectory data is normal, and the reference generation unit includes: A configuration in which the determination reference data is generated based on a combination of the plurality of trajectory data acquired by the acquisition unit and the determination result data may be adopted as a fifth aspect.

  According to the intersection monitoring system of the fifth aspect, determination reference data is generated based on a determination result indicating whether or not the movement of the moving object is normal in addition to the movement of the moving object indicated by the plurality of trajectory data. Therefore, compared with the case where the determination reference data generated based only on the movement of the moving object indicated by the plurality of trajectory data is used, a highly accurate determination result can be obtained.

In the intersection monitoring system according to any one of the first to fifth aspects, the moving image captured by the imaging device is converted into a moving image captured from an imaging angle different from the imaging angle of the imaging device, and the converted moving image A configuration including a trajectory generation unit that generates the trajectory data from an image may be employed as the sixth aspect.

In the intersection monitoring system according to any one of the first to fifth aspects, a moving image captured from an imaging angle different from an imaging angle of the imaging device indicates a trajectory of the moving body indicated by the moving image captured by the imaging device. A configuration in which a trajectory generating unit that converts the trajectory of the moving body and generates the trajectory data indicating the trajectory after conversion may be employed as the seventh aspect.

According to the intersection monitoring system of the sixth or seventh aspect, the trajectory of the moving object indicated by the moving images taken from different imaging angles can be handled as the trajectory of the moving object indicated by the moving images taken from the same imaging angle. it can. Therefore, it is not necessary to prepare a criterion for each of different imaging angles, and labor required for creating a criterion and checking a determination result is reduced.

In the intersection monitoring system according to the sixth or seventh aspect, the configuration is such that the acquisition unit acquires imaging angle data of the imaging device, and the trajectory generation unit performs the conversion based on the imaging angle data. You may employ | adopt as an aspect of 8 .

According to the intersection monitoring system of the eighth aspect, the trajectory of the moving body is converted based on the imaging angle of the imaging device. Therefore, the trouble of the user converting the locus of the moving body is reduced.

In the intersection monitoring system according to the sixth or seventh aspect, the acquisition unit acquires coordinate data of three or more points in the still image of the intersection captured by the imaging device, and the trajectory generation unit includes the coordinate data A configuration in which the conversion is performed based on the above may be adopted as the ninth aspect.

According to the intersection monitoring system of the ninth aspect, the trajectory of the moving object is converted based on the coordinate data of three or more points in the still image at the intersection. Therefore, the trouble of the user converting the locus of the moving body is reduced.

The figure which showed the whole structure of the intersection monitoring system concerning one Embodiment. The figure which showed the basic composition of the computer used for the terminal device and server apparatus concerning one Embodiment. The figure which showed the function structure of the terminal device concerning one Embodiment. The figure which illustrated the data composition of the judgment standard database memorized by the terminal unit and server apparatus concerning one embodiment. The figure which illustrated the data structure of the moving image database memorize | stored in the terminal device and server apparatus concerning one Embodiment. The figure which illustrated the data composition of the locus database memorized by the terminal unit and server apparatus concerning one embodiment. The figure which showed the function structure of the server apparatus concerning one Embodiment. The figure which showed operation | movement of the terminal device concerning one Embodiment. The figure which showed operation | movement of the server apparatus concerning one Embodiment. The figure which showed operation | movement of the terminal device concerning one Embodiment. The figure which showed the moving image browsing page displayed on the terminal device concerning one Embodiment. The figure which showed operation | movement of the server apparatus concerning one Embodiment. The figure which showed operation | movement of the server apparatus concerning one Embodiment. The figure which showed operation | movement of the terminal device concerning one Embodiment.

[Embodiment]
Below, the intersection monitoring system 1 concerning one Embodiment of this invention is demonstrated. FIG. 1 is a diagram showing an overall configuration of an intersection monitoring system 1 for monitoring the movement of a moving body such as a vehicle or a passerby that passes through the intersection 5, and a traffic light installed at the intersection 5.

  The intersection 5 illustrated in FIG. 1 is a standard crossroad with one lane on one side (hereinafter referred to as “one lane standard crossroad”). The type of the intersection 5 is not limited to the one-sided single lane standard crossroad, and may be any type of existing intersection.

  At the intersection 5 illustrated in FIG. 1, four traffic lights 6 are installed as vehicle traffic lights. In addition, eight traffic lights 7 are installed as traffic lights for pedestrians at the intersection 5 illustrated in FIG. A signal control device 8 is installed at the intersection 5. The signal control device 8 is connected to each of the traffic light 6 and the traffic light 7 by wire, for example, and controls the signal display state of each of the traffic light 6 and the traffic light 7. The signal control device 8 and each of the traffic light 6 and the traffic light 7 may be wirelessly connected.

  In addition, eight imaging apparatuses 11 are installed at the intersection 5 illustrated in FIG. As shown in FIG. 1, each of the eight imaging devices 11 captures images from different positions in different directions, and a moving image showing the movement of a moving body (vehicle, pedestrian, etc.) passing through the intersection 5. Generate data.

  The number and arrangement of the traffic light 6, the traffic light 7 and the imaging device 11 illustrated in FIG. 1 are merely examples, and the traffic light 6, the traffic light 7 and the imaging device are selected according to the type, size, tree arrangement, etc. of the intersection 5. The number and arrangement of 11 vary variously.

  A terminal device 12 is installed at the intersection 5. The terminal device 12 is connected to each of the imaging devices 11 by, for example, a wired connection, receives moving image data generated by the imaging device 11, and determines whether or not the movement of the moving object indicated by the moving image data is normal. (Hereinafter referred to as “moving body behavior determination”). Note that each of the terminal device 12 and the imaging device 11 may be wirelessly connected.

  The terminal device 12 is connected to the signal control device 8 by, for example, a wire, and receives various data (described later) used for moving body behavior determination from the signal control device 8, and also receives moving image data and movement received from the imaging device 11. Determination result data indicating the result of the body behavior determination is transmitted to the signal control device 8. The terminal device 12 and the signal control device 8 may be wirelessly connected.

  The signal control device 8 is connected to a server device 13 disposed in a traffic control center, for example, by wire or wirelessly. The location of the server device 13 is not limited to the traffic control center. The signal control device 8 transmits the moving image data and the determination result data received from the terminal device 12 to the server device 13.

  In addition to the intersection 5 illustrated in FIG. 1, the server device 13 receives moving image data, determination result data, and the like transmitted from the signal control device 8 installed at each of a number of different intersections 5. accumulate. In addition, the server device 13 generates determination criterion data indicating a criterion for determining the moving body behavior using the accumulated data, and transmits the determination criterion data to the terminal devices 12 installed at various intersections 5.

  The server device 13 can perform data communication with the terminal device 14 by wire or wireless. The terminal device 14 is, for example, a terminal device used by a person in charge who monitors the movement of a moving body that passes through the intersection 5. The server device 13 has, for example, a Web server function, and generates a Web page that displays the requested moving image data, determination result data, and the like in response to a request from the terminal device 14 and distributes the Web page to the terminal device 14.

  As described above, the terminal device 14 is, for example, a terminal device used by a person in charge of monitoring the movement of the moving body that passes through the intersection 5. The terminal device 14 has a browser function, for example, and has a browser function according to the user's operation. A web page is requested from the server device 13 and the web page distributed from the server device 13 is displayed.

  Among the components shown in FIG. 1, the intersection 5, the traffic light 6, the traffic light 7, and the signal control device 8 are already known, and thus description thereof is omitted. Moreover, among the components shown in FIG. 1, the imaging device 11, the terminal device 12, the server device 13, and the terminal device 14 constitute the intersection monitoring system 1. Among the components of the intersection monitoring system 1, the imaging device 11 and the terminal device 14 are known, and thus description thereof is omitted. Therefore, the terminal device 12 and the server device 13 will be described below.

  The terminal device 12 may be configured as a dedicated device using an ASIC or the like, or may be realized by a general computer. The server device 13 may also be configured as a dedicated device using an ASIC or the like, or may be realized by a general computer. Hereinafter, as an example, a case where both the terminal device 12 and the server device 13 are realized by a general computer will be described.

  FIG. 2 is a diagram showing a basic configuration of the computer 20 used for the terminal device 12 and the computer 30 used for the server device 13. The computer 20 outputs control data instructing imaging control to each of the memory 201 that stores various data, the processor 202 that performs various data processing according to the programs stored in the memory 201, and the imaging device 11, and An input / output IF 203 that is an interface for receiving input of moving image data from each of them and a communication IF 204 that is an interface for transmitting and receiving various data to and from the signal control device 8 are provided.

  The computer 30 is a memory 301 that stores various data, a processor 302 that performs various data processing in accordance with a program stored in the memory 301, a communication that is an interface that transmits and receives various data between the signal control device 8 and the terminal device 14. IF 303 is provided.

  FIG. 3 is a diagram illustrating a functional configuration of the terminal device 12. That is, when the computer 20 executes the program stored in the memory 201, the computer 20 operates as the terminal device 12 including the functional configuration unit illustrated in FIG. Below, the function structure part of the terminal device 12 is demonstrated.

  The moving image acquisition unit 1201 acquires moving image data from the imaging device 11. A trajectory generation unit 1202 (an example of a trajectory acquisition unit) generates trajectory data indicating the trajectory of the moving object indicated by the moving image data. Since the process of generating the trajectory data from the moving image data performed by the trajectory generation unit 1202 is known, the description thereof is omitted.

  Present state acquisition means 1203 acquires present state data indicating the present states of the traffic lights 6 and 7 from the signal control device 8. The timing specifying unit 1204 (an example of timing acquisition unit) specifies the timing at which the moving object that has moved along the trajectory indicated by the trajectory data has passed through the intersection 5 in the step of the current level indicated by the current status data.

  The storage unit 1205 stores various data. The data stored in the storage unit 1205 includes the type data indicating the type of the intersection 5 and the type of the current level, the current state data acquired from the signal control device 8 by the current state acquisition unit 1203, and the type of the intersection 5. And criteria data indicating criteria for determining the moving body behavior according to the type of the current floor, moving image data acquired from the imaging device 11 by the moving image acquisition means 1201, trajectory data generated by the trajectory generation means 1202, movement The determination result data indicating the result of the body behavior determination, imaging angle data indicating each imaging angle of the imaging device 11, and the like are included.

  The type data is data indicating the type of the intersection 5 where the terminal device 12 is installed, and the type of the current floor used by the signal control device 8 for display control of the traffic lights 6 and 7 at the intersection 5. The type data is data input to the terminal device 12 by an installer or the like when the terminal device 12 is installed, for example. However, for example, when the signal control device 8 changes the type of the displayed floor used in accordance with an instruction from the traffic control center or the like, the terminal device 12 receives a notification of the type of the currently used floor used from the signal control device 8, Update type data.

  The determination reference data is data generated according to the type of the intersection 5 in the server device 13 and is data received by the terminal device 12 from the server device 13. FIG. 4 is a diagram illustrating a data configuration of a determination criterion database in which the terminal device 12 stores determination criterion data. The determination reference database is a collection of records corresponding to each imaging direction, and includes [imaging direction], [first step], [second step],..., [I step] as fields. Note that “i” is the number of steps included in the current level according to the determination criterion data.

  In the “imaging direction”, data such as “first direction (right)” indicating the imaging direction of the moving image at the intersection 5 is stored. In each of [first step], [second step],..., [I-th step], determination reference data corresponding to each step of the current level is stored.

  For example, “J001_P001_S01_D01R_T1.crt” illustrated in FIG. 4 is “D01R” of the step identified by “S01” of the intersection type identified by “J001” and the current ladder type identified by “P001”. The determination reference data according to the period identified by “T1” in the imaging direction identified by “”.

  The determination criterion database illustrated in FIG. 4 includes determination criterion data corresponding to the period “T1” and the period “T1” in each of the [first step], [second step],. The criterion data corresponding to “T2” is stored. The period “T1” is a period (first period) within a predetermined time (for example, 10 seconds) immediately after the start of the step. The period “T2” is a period (second period) after the elapse of the period “T1” in the step. In other words, the criterion data database illustrated in FIG. 4 stores criterion data corresponding to each of two periods divided within a step. Note that the number of determination reference data corresponding to each step, that is, the number of periods for dividing each step is not limited to two.

  FIG. 5 is a diagram illustrating a data structure of a moving image database that stores moving image data received by the terminal device 12 from the imaging device 11. The moving image database is a collection of records corresponding to each of the moving image data divided at the timing when the present state (step of the present step) changes, and includes [intersection ID], [intersection type], [ It includes: “presented floor type”, “presented state”, “imaging direction”, “imaging start time”, “imaging end time”, and “moving image data”. In these fields, data corresponding to the field names is stored.

  FIG. 6 is a diagram illustrating a data configuration of a trajectory database that stores trajectory data generated by the terminal device 12. The trajectory database also stores determination result data indicating the result of the mobile device behavior determination performed by the terminal device 12 based on the trajectory data. The trajectory database is a collection of records corresponding to each of the moving objects that have passed through the intersection 5, and the fields include [moving object ID], [intersection ID], [intersection type], [presented ladder type], [presented state]. ], [Imaging direction], [passing start time], [passing end time], [passing start time in step], [passing end time in step], [determination result], and [trajectory data]. In these fields, data corresponding to the field names is stored.

  The imaging angle data is data indicating the imaging angle of the imaging device 11, for example, data input to the terminal device 12 by an installer or the like when the imaging device 11 is installed. The imaging angle data is, for example, a combination of data indicating a vector from the reference position (for example, the center point of the intersection 5) toward the installation position of the imaging device 11 and data indicating the imaging direction of the imaging device 11. However, the format of the imaging angle data is not limited to this.

  Returning to FIG. 3, the description of the functional components of the terminal device 12 will be continued. The transmission unit 1206 transmits various data such as moving image data, trajectory data, and determination result data to the server device 13. The receiving unit 1207 receives various data such as determination reference data from the server device 13.

  The type acquisition unit 1208 reads the type data from the storage unit 1205. The transmission unit 1206 generates a request for determination criterion data including the type data read by the type acquisition unit 1208 and transmits the request to the server device 13. The receiving unit 1207 receives the determination reference data (determination reference data corresponding to the intersection type and the current floor type indicated by the type data) transmitted from the server device 13 in response to the request.

  The imaging angle acquisition unit 1209 reads imaging angle data from the storage unit 1205. The trajectory generation unit 1202 responds to the trajectory specified from the moving image according to the difference between the imaging angle of the imaging device 11 indicated by the imaging angle data read by the imaging angle acquisition unit 1209 and the imaging angle serving as a predetermined reference. The trajectory data indicating the trajectory after the coordinate conversion is performed is generated. Accordingly, the trajectory data generated by the trajectory generation unit 1202 indicates a trajectory from which the influence of the imaging angle of each imaging device 11 is removed.

  The determination criterion acquisition unit 1210 relates to each of the trajectory data generated by the trajectory generation unit 1202, and the determination reference data corresponding to the trajectory data attributes (imaging direction, current state, timing within the step, etc.) from the storage unit 1205. Is read. The determination unit 1211 determines whether or not the trajectory of the moving object indicated by the trajectory data generated by the trajectory generation unit 1202 indicates normal movement according to the determination reference data read by the determination reference acquisition unit 1210. In addition, the result of the determination performed by the determination unit 1211 is stored in [determination result] of the trajectory database (FIG. 6). The above is the description of the functional configuration unit of the terminal device 12.

  FIG. 7 is a diagram showing a functional configuration of the server device 13. That is, when the computer 30 executes the program stored in the memory 301, the computer 30 operates as the server device 13 including the functional configuration unit illustrated in FIG. Hereinafter, a functional configuration unit of the server device 13 will be described.

  The receiving unit 131 receives various data from the terminal device 12 and the terminal device 14. Specifically, the receiving unit 131 receives moving image data, trajectory data, determination result data, and the like from the terminal device 12. The receiving unit 131 receives a request for moving image data and determination result data from the terminal device 14. Furthermore, the receiving unit 131 receives determination result data indicating a result of determining whether or not the movement of the moving object is normal by the user of the terminal device 14 viewing the moving image.

  The storage unit 132 stores various data. Specifically, the storage unit 132 stores a determination reference database (FIG. 4), a moving image database (FIG. 5), a trajectory database (FIG. 6), and the like, similar to the storage unit 1205 of the terminal device 12. However, the criterion data stored in the storage unit 132 stores criterion data corresponding to various combinations of intersection types and current floor types. In addition, moving image data transmitted from various terminal devices 12 is stored in the moving image database stored in the storage unit 132. The trajectory database stored by the storage unit 132 stores trajectory data and determination result data transmitted from various terminal devices 12.

  The trajectory acquisition unit 133 reads a large number of trajectory data from the storage unit 132. The reference generation unit 134 identifies outliers of these trajectory data based on a large number of trajectory data read by the trajectory acquisition unit 133, and generates determination reference data indicating the characteristics of the trajectory data identified as outliers. . As a method for the reference generating unit 134 to specify the outlier of the trajectory data, for example, an unsupervised learning method such as a well-known One Class Support Vector Machine is used.

  The determination result acquisition unit 135 reads out the determination result data (determination result data indicating the determination result by visual observation of the user of the terminal device 14) received by the server device 13 from the terminal device 14 from the storage unit 132. The reference generation unit 134 generates determination reference data based on the determination result data read by the determination result acquisition unit 135. As a method for generating the determination reference data by using the determination result data by the reference generation unit 134, for example, a known supervised learning method such as Support Vector Machine is used. In this case, the determination result data is used as teacher data.

  The determination unit 136 determines whether or not the trajectory of the moving object indicated by the trajectory data read by the trajectory acquisition unit 133 indicates normal movement according to the determination reference data generated by the reference generation unit 134.

  The transmission unit 137 transmits various data to the terminal device 12 and the terminal device 14. Specifically, the transmission unit 137 transmits the determination reference data to the terminal device 12. The transmission unit 137 transmits moving image data and determination result data to the terminal device 14. The above is the description of the functional configuration unit of the server device 13.

  Subsequently, the operation of the intersection monitoring system 1 will be described. In the intersection monitoring system 1, the moving body behavior determination is performed based on the determination criteria according to the type of the intersection to be monitored and the type of the current floor used at the intersection to be monitored. Accordingly, when monitoring by the intersection monitoring system 1 is newly started at a certain intersection, if the determination criteria according to the type of this intersection and the type of the present floor used at this intersection are not yet prepared, Preparations for the criteria are made.

For example, one of the following two methods is employed in the determination criterion preparation phase in the intersection monitoring system 1.
(Unsupervised learning method) A method for generating criterion data by unsupervised learning using collected trajectory data.
(Supervised learning method) A method of generating determination reference data by supervised learning using determination result data indicating the result of visual determination.

  The operation of the intersection monitoring system 1 in the preparation phase when the unsupervised learning method is adopted will be described below. In the preparation phase when the unsupervised learning method is adopted, the terminal device 12 generates a predetermined number of pieces of trajectory data and transmits them to the server device 13. Further, the server device 13 generates determination reference data by unsupervised learning using a predetermined number of trajectory data transmitted from the terminal device 12.

  FIG. 8 is a diagram illustrating the operation of the terminal device 12 in the preparation phase when the unsupervised learning method is employed. First, the terminal device 12 assigns an initial value “0” to a counter h for counting the number of locus data to be generated (step S101). Subsequently, the terminal device 12 receives moving image data sequentially transmitted from each of the plurality of imaging devices 11 (step S102). At the same time, the terminal device 12 receives the display state data sequentially transmitted from the signal control device 8 (step S103). The terminal device 12 divides the moving image data at the switching timing of the steps of the display step indicated by the display state data, and stores the divided moving image data in the moving image database (FIG. 5) of the storage unit 1205 ( Step S104). Note that the processing in steps S102 to S104 regarding the subsequent moving image data is continued in parallel with the processing after the following step S105 regarding the preceding moving image data.

  The terminal device 12 identifies the moving object indicated by the moving image data stored in step S104, and generates locus data indicating the locus of the identified moving object (step S105). Hereinafter, the number of moving objects specified in step S105, that is, the number of trajectory data generated in step S105 is n. The terminal device 12 performs a coordinate conversion process on each of the generated trajectory data (step S106). This coordinate conversion process is a process of removing standardized trajectory data by removing the influence of the difference in imaging angle of each imaging device 11. The terminal device 12 stores the trajectory data subjected to the coordinate conversion process in the trajectory database (FIG. 6) of the storage unit 1205 (step S107). The terminal device 12 adds the number n of trajectory data newly stored in step S107 to the counter h (step S108).

  The terminal device 12 determines whether or not the counter h has reached a predetermined number (10000 in the example of FIG. 8) (step S109). When the counter h has not reached the predetermined number (step S109; No), the terminal device 12 continues the processing of steps S102 to S108. When the counter h has reached the predetermined number (step S109; Yes), the terminal device 12 stops the processing of steps S102 to S108, and stores the stored moving image data and trajectory data with their attributes (imaging direction, current indication). The data indicating the status etc. is transmitted to the server device 13 (step S110).

  FIG. 9 is a diagram showing the operation of the server device 13 in the preparation phase when the unsupervised learning method is adopted. The server device 13 receives the moving image data and the trajectory data transmitted by the terminal device 12 in step S110 described above together with data indicating their attributes (imaging direction, current state, etc.) (step S201). The server device 13 stores the received data in the moving image database (FIG. 5) or the trajectory database (FIG. 6) of the storage means 132 (step S202).

  Subsequently, the server device 13 groups the trajectory data stored in step S202 for each step of the current level (step S203). In the example of FIG. 9, the number of steps of the current floor is 10. Accordingly, in step S203, the server device 13 classifies the trajectory data stored in step S202 into 10 groups corresponding to each of the first to tenth steps.

  Subsequently, the server device 13 substitutes an initial value “1” for a counter j indicating the order of steps in the current floor (step S204). The server device 13 groups the trajectory data of the jth step for each period within the step (step S205). In the example of FIG. 9, the step is divided into two periods (a period immediately after the start of the step and a period thereafter). Accordingly, in step S205, the server apparatus 13 classifies the trajectory data of the j-th step into two groups corresponding to the first period and the second period in the step.

  Subsequently, the server device 13 substitutes an initial value “1” for the counter k indicating the order of the periods in the step (step S206). Subsequently, the server device 13 generates determination criterion data by unsupervised learning using the trajectory data of the kth period of the jth step (step S207). The server device 13 stores the generated determination criterion data in the determination criterion database (FIG. 4) of the storage unit 132 (step S208).

  Subsequently, the server device 13 determines whether or not the counter k has reached “2” which is the number of periods divided within the step (step S209). If the counter k has not reached “2” (step S209; No), the server device 13 increments the counter k by 1 (step S210), and returns the process to step S207.

  If it is determined in step S209 that the counter k has reached 2 (step S209; Yes), the server device 13 determines whether or not the counter j has reached “10”, which is the number of steps of the current floor ( Step S211). When the counter j has not reached “10” (step S211; No), the server device 13 increments the counter j by 1 (step S212), and returns the process to step S205. When the counter j has reached “10” (step S211; Yes), the server device 13 ends the series of processes. The above is the operation of the intersection monitoring system 1 in the preparation phase when the unsupervised learning method is adopted.

  Next, the operation of the intersection monitoring system 1 in the preparation phase when the supervised learning method is adopted will be described. In the preparation phase in the case where the supervised learning method is adopted, the terminal device 14 displays a moving image for the user and acquires the result of the user's visual determination. The server device 13 transmits moving image data to the terminal device 14. Further, the server device 13 generates determination reference data by supervised learning based on the result of visual determination by the user of the terminal device 14.

  FIG. 10 is a diagram showing the operation of the terminal device 14 in the preparation phase when the supervised learning method is adopted. First, the user of the terminal device 14 operates the terminal device 14 to display a browser screen, and inputs a predetermined URL (Uniform Resource Locator) to provide a Web page provided by the server device 13 (hereinafter, “moving image browsing page”). ”) (Step S301).

  FIG. 11 shows a moving image browsing page. Hereinafter, the operation of the terminal device 14 will be described with reference to FIGS. 10 and 11 together. On the moving image browsing page, an intersection ID input field for identifying an intersection to be monitored which the user wants to view the moving image is displayed. When the user inputs an intersection ID in this input field, the terminal device 14 transmits a data transmission request corresponding to the intersection ID input by the user to the server device 13 (step S302). The terminal device 14 receives data data transmitted from the server device 13 as a response to the transmission request (step S303). The data received by the terminal device 14 in step S303 includes moving image data cut out for each moving object that has passed the intersection corresponding to the intersection ID input by the user, trajectory data corresponding to the moving image data, and the moving image. This is data indicating image data attributes (imaging direction, display state, etc.).

  The terminal device 14 uses the data received from the server device 13 to update the display of a list (hereinafter referred to as “moving image list”) included in the moving image browsing page. In the moving image list, the passage start time and the imaging direction regarding the moving body that has passed the intersection at the step selected by the user on the current level are displayed. Each row of the moving image list is associated with moving image data of the corresponding moving object. The list includes a determination result display column, and the result of the visual determination by the user is displayed in this column. Therefore, before the visual determination by the user is performed, the determination result display field is empty. Also, before the user selects a step, the first step is selected by default, and information regarding the moving object that has passed the intersection in the first step is displayed in the moving image list.

  A drop-down list for the user to select a step is displayed on the moving image browsing page. When the user selects any step from the drop-down list, information regarding the step selected by the user is displayed in the moving image list. When the user selects any row from the moving image list and clicks the “play” button or the like, the terminal device 12 acquires a moving image reproduction instruction for the moving object selected by the user (step S304). A moving image for which a reproduction instruction has been issued is reproduced in an area below the list (step S305). In step S <b> 305, the terminal device 12 displays the moving object's trajectory indicated by the trajectory data together with the moving image, for example, with an arrow or the like.

  The user visually observes the moving image reproduced on the moving image browsing page, determines whether or not the movement of the moving body is normal, and clicks the “normal” button or the “abnormal” button. In response to the user operation, the terminal device 14 acquires determination result data indicating the result of the visual determination by the user (step S306). The terminal device 14 temporarily stores the acquired determination result data and reflects the content of the acquired determination result data in the determination result display field of the moving image list.

  When the user wants to visually determine the movement of another moving body (step S307; No), the user selects an appropriate line from the moving image list, and then clicks the “play” button to visually check the moving image. When it is determined that the movement of the moving body is normal, the “normal” button is clicked. Thereby, the processing of steps S304 to S306 is repeated.

  On the other hand, when the user wishes to complete the visual confirmation of the moving image, the user clicks a “send” button or the like. In response to this user operation (step S307; Yes), the terminal device 14 transmits the determination result data acquired in step S306 to the server device 13 (step 308).

  FIG. 12 is a diagram illustrating an operation performed by the server device 13 when the user of the terminal device 14 performs the visual determination of the moving image in the preparation phase when the supervised learning method is employed. When the server apparatus 13 receives a transmission request for data transmitted from the terminal apparatus 14 in step S302 of FIG. 10 (step S401), the server apparatus 13 stores the data in the table included in the trajectory database (FIG. 6) stored in the storage unit 132. Then, a record whose [intersection ID] matches the intersection ID included in the transmission request is extracted (step S402).

  Subsequently, for each record extracted in step S402, the server device 13 searches for a corresponding record from the moving image database (FIG. 5) stored in the storage unit 132 (step S403). Specifically, the server apparatus 13 matches the record extracted from the trajectory database with [intersection ID], and indicates an imaging period that covers the passage period indicated by [passing start time] and [passing end time] [imaging start time] ] And [imaging end time] are searched from the moving image database.

  Subsequently, the server device 13 cuts out the portion corresponding to the passage period indicated by the [passing start time] and [passing end time] of the corresponding trajectory database record from the moving image data included in the record searched in step S403 ( Step S404). The server device 13 indicates the trajectory data included in the record extracted from the trajectory database in step S402 and the moving image data cut out in step S404 with respect to the trajectory data, and the attributes of these data (imaging direction, current state, etc.). The data is transmitted to the terminal device 14 together with the data (step S405).

  Thereafter, the server device 13 receives the determination result data transmitted from the terminal device 14 in step S308 of FIG. 10 (step S406), and updates the trajectory database (FIG. 6) of the storage unit 132 with the received determination result data (step S406). Step S407).

  FIG. 13 is a diagram illustrating an operation when the server device 13 generates the determination reference data in the preparation phase when the supervised learning method is employed. In addition, the server apparatus 13 produces | generates criterion data for every combination of an intersection classification and a presenting floor classification. The server device 13 performs the operation shown in FIG. 13 for each of the various combinations of intersection types and current floor types.

  First, the server device 13 records from the trajectory database (FIG. 6) a record corresponding to the combination of the intersection type and the current floor type for which the determination reference data is generated, and stores normality or abnormality in the determination result. The recorded record is extracted (step S501). If the number of records extracted in step S501 is less than the predetermined number (10000 in the example of FIG. 13) (step S502; No), the number of determination result data necessary for generating the determination reference data cannot be obtained, and the server The device 13 ends the process.

  When the number of records extracted in step S501 is equal to or greater than a predetermined number (step S501; Yes), the server device 13 groups the extracted records for each step of the current level (step S503).

  Subsequently, the server device 13 substitutes an initial value “1” for a counter j indicating the order of steps in the current floor (step S504). The server device 13 groups records corresponding to the jth step for each period within the step (step S505).

  Subsequently, the server device 13 substitutes an initial value “1” for the counter k indicating the order of the periods in the step (step S506). Subsequently, the server device 13 generates determination criterion data by supervised learning using a record corresponding to the k-th period of the j-th step (step S507). In step S507, the server device 13 uses the data stored in [determination result] as teacher data.

  The server device 13 updates the criterion database (FIG. 4) in the storage unit 132 with the generated criterion data (step S508). Subsequently, the server device 13 determines whether or not the counter k has reached “2” which is the number of periods divided within the step (step S509). If the counter k has not reached “2” (step S509; No), the server device 13 increments the counter k by 1 (step S510), and returns the process to step S507.

  If it is determined in step S509 that the counter k has reached 2 (step S509; Yes), the server device 13 determines whether or not the counter j has reached “10”, which is the number of steps of the current floor ( Step S511). When the counter j has not reached “10” (step S511; No), the server device 13 increments the counter j by 1 (step S512), and returns the process to step S505. When the counter j has reached “10” (step S511; Yes), the server device 13 ends the series of processes. The above is the operation of the intersection monitoring system 1 in the preparation phase when the supervised learning method is adopted.

  After the preparation phase, the preparation of the judgment criteria is completed, and the transition to the operation phase becomes possible. Hereinafter, the operation of the intersection monitoring system 1 in the operation phase will be described. FIG. 14 is a diagram illustrating the operation of the terminal device 12 in the operation phase. Prior to the operation phase, the terminal device 12 receives determination criterion data corresponding to the type of intersection indicated by the type data stored in the storage unit 1205 and the type of the current floor from the server device 13, and receives the received determination criterion data. It is stored in the judgment criterion database (FIG. 4) of the storage means 1205.

  In the above state, the terminal device 12 receives the moving image data sequentially transmitted from each of the plurality of imaging devices 11 (step S601). At the same time, the terminal device 12 receives the present state data sequentially transmitted from the signal control device 8 (step S602). The terminal device 12 divides the moving image data at the switching timing of the steps of the display step indicated by the display state data, and stores the divided moving image data in the moving image database (FIG. 5) of the storage unit 1205 ( Step S603). The processes in steps S601 to S603 related to the subsequent moving image data are continued in parallel with the processes in and after the following step S604 related to the preceding moving image data.

  The terminal device 12 identifies the moving object indicated by the moving image data stored in step S603, and generates locus data indicating the locus of the identified moving object (step S604). The terminal device 12 performs coordinate conversion processing on the generated trajectory data (step S605). The terminal device 12 stores the trajectory data subjected to the coordinate conversion process in the trajectory database (FIG. 6) of the storage unit 1205 (step S606).

  Subsequently, the terminal device 12 obtains determination reference data corresponding to the current state of the trajectory data stored in step S606 (the step of the current display step) and the in-step passage start time from the determination reference database (FIG. 4) of the storage unit 1205. Read (step S607). The terminal device 12 performs moving body behavior determination on the trajectory data stored in step S606 based on the read determination criterion data (step S608). The terminal device 12 stores the determination result data indicating the result of the moving body behavior determination in the trajectory database (FIG. 6) of the storage unit 1205 (step S609).

  Subsequently, the terminal device 12 transmits the record of the moving image database that stores the moving image data in step S603 and the record of the determination result database that stores the determination result data in step S609 to the server device 13 (step S610).

  The record of the moving image database transmitted from the terminal device 12 in step S610 is added to the moving image database (FIG. 5) of the server device 13. In addition, the record of the trajectory database transmitted from the terminal device 12 in step S610 is added to the trajectory database (FIG. 6) of the server device 13. Thereby, moving image data, trajectory data, and determination result data are accumulated in the server device 13. The above is the description of the operation of the intersection monitoring system 1 in the operation phase.

  The operation in the operation phase described above is performed between each of the terminal devices 12 installed at various intersections 5 and the server device 13, so that the moving image database of the server device 13 is captured at various intersections 5. In addition, the moving image data and the trajectory database and the determination result data relating to the moving body that has passed through the various intersections 5 are accumulated in the trajectory database of the server device 13. These data stored in the server device 13 are used, for example, for displaying a moving image browsing page (FIG. 11). Therefore, for example, the user of the terminal device 14 displays the moving image browsing page on the terminal device 14 and confirms the movement of the moving body determined to be abnormal with the moving image, thereby improving the arrangement of the traffic light 6 and the traffic light 7. And adjustment of signal display cycle, split, or offset.

  As described above, according to the intersection monitoring system 1, in the operation phase, it is determined whether or not the movement of the moving body passing through the intersection 5 is normal regardless of the human hand. Therefore, the user does not need to confirm the moving image in order to detect an abnormal movement of the moving body.

  Further, when the unsupervised learning method is adopted in the preparation phase, according to the intersection monitoring system 1, the determination reference data is generated regardless of human hands. In the intersection monitoring system 1, the determination reference data corresponding to each of the displayed states (steps of the displayed steps) is generated. Therefore, even if the determination reference data is generated regardless of the hand of a person, High reliability is obtained.

  Further, when the supervised learning method is adopted in the preparation phase, the intersection monitoring system 1 is more reliable than the case where the unsupervised learning method is adopted based on the result of the determination made by the user through the visual determination. Determination criteria data with higher performance is generated. In the intersection monitoring system 1, the moving image and the trajectory of the moving body corresponding to each of the displayed states (steps of the displayed steps) are displayed to the user, so that the user can easily make a visual determination. In addition, as in the case where the unsupervised learning method is adopted, even when the supervised learning method is adopted, the determination reference data is generated according to each of the present state (steps of the present step), Compared with the case where uniform determination reference data is generated regardless of the display state, highly reliable determination reference data is generated.

[Modification]
The embodiment described above can be variously modified. Examples of these modifications are shown below. Note that the above-described embodiment and the modifications described below may be combined as appropriate.

(1) At least a part of the processing performed by the terminal device 12 in the above-described embodiment may be performed by the server device 13. For example, the terminal device 12 may not perform the moving body behavior determination in the operation phase, the terminal device 12 may transmit the trajectory data or the like to the server device 13, and the movement pair behavior determination may be performed in the server device 13. In addition, at least a part of the processing performed by the server device 13 in the above-described embodiment may be performed by the terminal device 12. For example, the terminal device 12 may generate the determination reference data instead of the server device 13. Further, a configuration in which the terminal device 12 and the server device 13 are integrated into one device may be employed.

(2) In the above-described embodiment, in the preparation phase in the case where the supervised learning method is adopted, the teacher data used for generating the determination reference data is determination result data indicating the result of the visual determination of the user of the terminal device 14. is there. The data used as the teacher data in the preparation phase when the supervised learning method is adopted is not limited to the determination result data indicating the result of the visual determination by the user. For example, a configuration may be employed in which trajectory data indicating the movement of a moving body that has caused a traffic accident in the past is used as teacher data.

(3) In the above-described embodiment, all the moving image data and trajectory data generated by the imaging device 11 are stored in the terminal device 12 or the server device 13. May not be stored in the terminal device 12 or the server device 13. For example, a configuration may be employed in which moving image data in which a moving body that has been determined to be abnormal is not captured, or trajectory data generated from the moving image data is not stored in the terminal device 12 or the server device 13. .

(4) In the above-described embodiment, the coordinate conversion process for removing the influence due to the difference in the imaging angle is performed based on the imaging angle data indicating the imaging angle of the imaging device 11. Instead, for example, the installer of the terminal device 12 designates three or more feature points of the intersection 5 (for example, the intersection of the road center line and the stop line) shown in the still image captured by the imaging device 11. A configuration in which the terminal device 12 performs coordinate conversion processing based on the specified coordinates in the still image of these three points may be employed. In this case, the trajectory data is such that the coordinates of the feature points designated by the installer of the imaging device 11 move to the coordinates of the same kind of feature points in the still image captured at the reference imaging angle. The coordinate transformation is performed.

(5) In the above-described embodiment, the coordinate conversion process for removing the influence due to the difference in the imaging angle is performed on the trajectory data. Alternatively, coordinate conversion processing may be performed on the moving image data. In this case, the trajectory data is generated based on the moving image data on which the coordinate conversion process has been performed.

(6) In the above-described embodiment, the determination reference data is generated by the server device 13 regardless of human hands. Instead, determination criterion data created by an administrator of the intersection monitoring system 1 or the like may be used. Even in this case, the manager of the intersection monitoring system 1 only needs to create the determination reference data for each step of the displayed floor, so that it is compared with the case where the determination reference data is generated without considering the current display state. And labor is reduced.

(7) A configuration in which the criterion data is updated in the operation phase may be adopted. For example, in the operation phase, when the user of the terminal device 14 is browsing the moving image and the trajectory of the moving body on the moving image browsing page (FIG. 11), an error in the determination result by the terminal device 12 may be found. In this case, the user clicks the “send” button after clicking the “normal” button or “abnormal” button corresponding to the correct determination result. As a result, the determination result data indicating the result of the visual determination of the user is reflected in the trajectory database (FIG. 6) of the server device 13. For example, the server device 13 generates determination criterion data using [determination result] stored in a trajectory database updated every elapse of a predetermined time as teacher data, and the determination criterion data (FIG. 4) is generated using the generated determination criterion data. ). Thereafter, if the terminal device 12 receives the updated determination reference data from the server device 13 and uses it for the mobile body behavior determination, the reliability of the determination result is improved.

(8) In the above-described embodiment, as illustrated in FIG. 4, when a plurality of imaging devices are installed at one intersection, determination reference data is prepared for each imaging direction of the imaging devices. However, the same determination reference data may be used for two or more imaging directions that are symmetrical to each other at a symmetric intersection such as a one-lane standard crossroad.

DESCRIPTION OF SYMBOLS 1 ... Intersection monitoring system, 5 ... Intersection, 6 ... Traffic light, 7 ... Traffic light, 8 ... Signal control device, 11 ... Imaging device, 12 ... Terminal device, 13 ... Server device, 14 ... Terminal device, 20 ... Computer, 30 ... Computer 131 ··· Receiving means 132 ··· Storage means 133 ··· Trajectory acquiring means 134 ··· Reference generating means 135 ··· Determination result acquiring means 136 ··· Determination means 137 ··· Sending means 201 · · · Memory · 202 ··· Processor 203 ... Input / output IF, 204 ... Communication IF, 301 ... Memory, 302 ... Processor, 303 ... Communication IF, 1201 ... Moving image acquisition means, 1202 ... Track generation means, 1203 ... Current state acquisition means, 1204 ... Timing specification means, 1205 ... Storage means, 1206 ... Transmission means, 1207 ... Reception means, 1208 ... Type acquisition means, 1209 ... Imaging ang Acquisition means, 1210 ... criterion acquisition unit, 1211 ... judging means

Claims (9)

And locus data of a moving body installed an imaging device at an intersection passes through the intersection indicated by the moving image captured, the type of the intersection, before at the time of classification and the passage of current示階ladder of the traffic signal of the intersection Kigen Acquisition means for acquiring determination reference data indicating a reference for moving body behavior determination according to a combination of steps of the indicated steps;
An intersection monitoring system comprising: a determination unit that determines whether the movement indicated by the trajectory data is normal in light of the determination reference data. The acquisition means acquires timing data indicating the timing at which the moving body passes the intersection in the step of the current level of the traffic light,
The intersection monitoring system according to claim 1, wherein the acquisition unit acquires the determination reference data according to the timing data. The acquisition means acquires display state data indicating a display state of a traffic signal at the intersection at a time when a moving body passes the intersection.
The intersection monitoring system according to claim 1, wherein the acquisition unit acquires the determination reference data according to the display state data. The acquisition means acquires a plurality of trajectory data,
The intersection monitoring system according to any one of claims 1 to 3, further comprising reference generation means for generating the determination reference data based on the plurality of trajectory data acquired by the acquisition means. The acquisition means acquires determination result data indicating whether or not the moving body indicated by each of the plurality of trajectory data is normal,
The intersection monitoring system according to claim 4, wherein the reference generation unit generates the determination reference data based on a combination of the plurality of trajectory data acquired by the acquisition unit and the determination result data. 2. A trajectory generation unit configured to convert a moving image captured by the imaging apparatus into a moving image captured from an imaging angle different from an imaging angle of the imaging apparatus and generate the trajectory data from the converted moving image. The intersection monitoring system according to any one of 1 to 5 . The trajectory of the moving object indicated by the moving image captured by the imaging device is converted into the trajectory of the moving object indicated by the moving image captured from an imaging angle different from the imaging angle of the imaging device, and the trajectory indicating the converted trajectory The intersection monitoring system according to any one of claims 1 to 5 , further comprising trajectory generating means for generating data. The acquisition means acquires imaging angle data of the imaging device,
The intersection monitoring system according to claim 6 or 7 , wherein the trajectory generation means performs the conversion based on the imaging angle data. The acquisition means acquires coordinate data of three or more points in the still image of the intersection imaged by the imaging device,
The intersection monitoring system according to claim 6 or 7 , wherein the trajectory generating means performs the conversion based on the coordinate data.