JP2019003329A - Information processor, information processing method, and program - Google Patents

Abstract

To make it possible to create an identification model using proper data having less errors.SOLUTION: A data storage part (102) stores event data containing a feature quantity of an event of an object created in advance. A video acquisition part (101) acquires a video. A feature quantity creation part (103) creates a feature quantity of the event of the object in the acquired video. A data selection part (105) selects event data containing a feature quantity similar to the feature quantity created by the feature quantity creation part (103) from the event data stored in the data storage part (104). An identification model creation part (106) creates an identification model identifying an event of the object in the video by using the feature quantity of the selected event data.SELECTED DRAWING: Figure 1

Description

  The present invention relates to an information processing apparatus, an information processing method, and a program for generating an identification model used for identifying an object in video data.

  Conventionally, a method for creating an identification model for identifying an event of an object from video data (identifying the object and the state of the object) has been proposed. Since learning data is required to create this identification model, it takes time to collect learning data. Also, it is difficult to check whether the collected learning data is sufficient.

  On the other hand, in Non-Patent Document 1, a model to be detected is prepared in advance, information such as a camera position and a background image is input as environment information, and a learning sample specialized for the environment is generated by CG (computer graphic). There is a way to do additional learning. Thereby, the cost of creating learning data in accordance with the environment is reduced.

Narumi Tsuchiya, Atsushi Yamauchi, Takayoshi Yamashita, Hironobu Fujiyoshi, Learning Efficiency in Object Detection by Hybrid Type Transfer Learning, IEICE Tech. 112, no. 385, PRMU2012-122. pp. 329-334, January 2013

  However, in the technique of Non-Patent Document 1, there is a possibility that data representing an event (object state) of an object that does not actually occur in a set scene (scene) is generated by CG. For this reason, unnecessary data is mixed into the learning data, which may cause undetected / misdetected data during identification.

  Therefore, an object of the present invention is to enable generation of an identification model that can accurately identify an event of an object in a scene.

  The present invention provides a storage means for storing a plurality of event data including a feature amount of an event of an object generated in advance, an image acquisition means for acquiring an image, and an event of an object in the acquired image. A feature amount creating unit that creates a feature amount, and a selection that selects event data including a feature amount similar to the feature amount created by the feature amount creating unit from among the event data stored in the storage unit And a model creating means for creating an identification model for identifying the event of the object in the video using the feature amount of the selected event data.

  According to the present invention, it is possible to generate an identification model that can accurately identify an event of an object in a scene.

It is a schematic block diagram of the information processing apparatus of 1st Embodiment. It is a figure which shows the example of input of the normal action of a target object in 1st Embodiment. It is a figure which shows an example of the action data of normal action. It is explanatory drawing of the data search using a hash function group. It is a figure which shows the example of a synthesis | combination of the collected action data and an image | video. It is explanatory drawing of the example of registration to a data storage part. It is a flowchart of the process of the information processing apparatus of 1st Embodiment. It is a schematic block diagram of the information processing apparatus of 2nd Embodiment. It is a figure which shows an example of the action data of abnormal action. It is explanatory drawing of the example of input of the action data by label selection. It is a flowchart of a process of the information processing apparatus of 2nd Embodiment. It is a schematic block diagram of the information processing apparatus of 3rd Embodiment. It is explanatory drawing of the example of action data input using map information. It is a flowchart of the process of the information processing apparatus of 3rd Embodiment.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings.
<First Embodiment>
FIG. 1A shows a schematic configuration example of the information processing apparatus 100 according to the first embodiment. As will be described later, the information processing apparatus 100 according to the present embodiment designates an event of an object that occurs in a video scene, and stores the event data of the designated object from a database that stores event data of the object created in advance. An event model similar to an event is selected to create an identification model. Further, in the information processing apparatus 100 according to the first embodiment, when creating an identification model, event data relating to a normal event of an object is collected. As a normal event of an object, for example, when a pedestrian walking on a pedestrian crossing in the video scene of an intersection is used as an object, the pedestrian's general pedestrian walks on a pedestrian crossing. You can list actions. Of course, this is an example, and the normal phenomenon of the object is not limited to the behavior of a pedestrian walking on a pedestrian crossing. In the first embodiment, description is given using an image obtained by photographing an outdoor intersection as an example of an image scene. In addition, an image scene is a public facility such as a commercial facility, a hospital, a nursing facility, or a station. It may be a scene inside the facility or its surroundings.

Hereinafter, in the information processing apparatus 100 of the present embodiment, a configuration and processing for generating an identification model by specifying a normal event of an object in a video scene will be described.
An information processing apparatus 100 illustrated in FIG. 1A includes a video acquisition unit 101, an input unit 102, a feature amount creation unit 103, a data storage unit 104, a data selection unit 105, an identification model creation unit 106, The identification model storage unit 107 and the display unit 108 are included.

The video acquisition unit 101 acquires, for example, video data to be monitored captured by a monitoring camera or the like installed at an intersection or public facility, and outputs the acquired video data to the display unit 108 and the feature amount generation unit 103. .
FIG. 2 shows a display example in which the video data acquired by the video acquisition unit 101 is displayed on the screen of the display unit 108. FIG. 2 shows an example in which images 201-1 to 201-n for consecutive n frames among the images of the surveillance camera installed at the intersection are displayed on the screen of the display unit 108. It is assumed that the images 201-1 to 201-n for n frames illustrated in FIG. 2 show a pedestrian 221 walking on a crosswalk at an intersection. The frames 211-1 to 211-n and the attribute information list 212 in the screen illustrated in FIG. 2 will be described later.

  The input unit 102 acquires information such as an input instruction from a user via a GUI (graphical user interface) using the screen display of the display unit 108. That is, in the present embodiment, the user can input an instruction regarding a normal event of the target object via the input unit 102 while viewing the video displayed on the display unit 108. Hereinafter, as an example of the images 201-1 to 201-n shown in FIG. 2, an example in which the pedestrian 221 walks on the pedestrian crossing at the intersection will be described, and an instruction input example of a normal event of the object by the user will be described. To do.

  Here, when the action of the pedestrian 221 walking on the pedestrian crossing is designated as a normal event of the target object as shown in FIG. 2, the user designates the pedestrian 221 in the video via the input unit 102. An instruction is input to When an instruction is input from the user via the input unit 102, the information processing apparatus 100 sets a predetermined frame for the pedestrian 221 in the video based on the instruction input. As the instruction input by the user at this time, for example, an input method for instructing, for example, the upper left position and the lower right position of the object (pedestrian 221) on the video via the GUI can be used. When the user designates the upper left and lower right positions of the pedestrian 221 in the video via the input unit 102, the information processing apparatus 100 sets the designated positions as the upper left position and the lower right position of the frame. Set the rectangular frame to be used. The designation by the user and the frame setting method are not limited to this example, and other methods may be used.

  In the case of the present embodiment, the setting of the frame for the pedestrian 221 is performed for each of the images 201-1 to 201-n for consecutive n frames. As a result, frames 211-1 to 211-n are set for the images 201-1 to 201-n for the n frames, respectively. Note that the position is specified by the user only for the first one of the consecutive n frames, and for the subsequent 2 to n frames, the information processing apparatus 100 uses the known tracking technique described in Reference Document 1 below to detect the object. The frame may be automatically set by tracking. The tracking method of the object is not limited to the example of Reference 1, and other tracking methods may be used. As described above, the information processing apparatus 100 according to the present embodiment deals with information on the frames 211-1 to 211-n set for the videos 201-1 to 201-n (hereinafter referred to as region information). Acquired as one of information related to an event of an object.

  Reference 1: M.M. Isard and A.M. Blake, Condensation-conditional density propagation for visual tracking, International Journal of Computer Vision, vol. 29, no. 1, pp. 5-28, 1998.

  Further, the information processing apparatus 100 according to the present embodiment also acquires the attribute information of the target object as information related to the normal event of the target object together with the area information described above. Examples of the attribute information include category information representing an object, environmental information such as weather, time information such as time and time zone, and the like. The information processing apparatus 100 according to the present embodiment can acquire information by inputting an instruction from the user via the input unit 102 for the attribute information.

  The attribute information list 212 illustrated in FIG. 2 is used when the user specifies attribute information via the input unit 102. The information processing apparatus 100 according to the present embodiment displays an attribute information list 212 as illustrated in FIG. 2 on the screen, and acquires attribute information specified by the user via the input unit 102 from the attribute information list 212. . The attribute information list 212 illustrated in FIG. 2 includes object category information (for example, category information representing a moving object such as a pedestrian or a bicycle), environmental information representing weather, time information representing time, time zone, and the like. The pull-down list is selectable by the user. Therefore, the user can designate attribute information by operating the pull-down list of the attribute information list 212 via the input unit 102 while viewing the videos 201-1 to 201-n in FIG. Note that the attribute information list 212 in FIG. 2 includes a pull-down list for designating the type of object, which will be described later. The attribute information is not limited to the information listed in the attribute information list 212 in FIG. 2, and other attribute information may be specified. The information processing apparatus 100 according to the present embodiment acquires attribute information specified by the user from the attribute information list 212 as one piece of information related to a normal event of the target object.

  Then, the information processing apparatus 100 according to this embodiment includes the area information acquired as described above (area information in the frames 211-1 to 211-n in the example of FIG. 2) and attribute information (attribute information in the example of FIG. 2). Attribute information specified in the list 212) is sent to the feature quantity creation unit 103.

  The feature amount creation unit 103 creates a feature amount in the event of the object in the video from the video data. For example, the feature quantity creation unit 103 calculates a motion vector of an object from video data, generates a feature vector having each element as an average value of the motion vector of the object, and creates the feature vector as a feature quantity To do. In the case of the example in FIG. 2, the feature amount creation unit 103 calculates a motion vector from the images 201-1 to 201-n for the frame set for the pedestrian 221, and calculates an average value of the motion vector for each element. The feature vector is created as a feature quantity. Further, the feature quantity creation unit 103 may obtain a feature quantity by, for example, HOF (Histogram of Optical Flow), MHOF (Multi Histogram of Optical Flow) shown in Reference Document 2 below. In addition, in HOF and MHOF, a feature quantity obtained as a histogram is obtained by dividing motion vectors into directions and adding intensities. In addition, for example, the feature amount creation unit 103 may obtain HOG (Histogram of Oriented Gradients) in which the gradient strength of appearance shown in Reference Document 3 below is a histogram for each direction, or other feature amounts. The feature amounts in the present embodiment are not limited to those according to these described methods. Then, the feature quantity creation unit 103 sends the feature quantity created as described above and the attribute information described above to the data selection unit 105 as event data relating to the normal event of the object in the acquired video. .

Reference 2: J. Pers, et al, Histograms of Optical Flow for Efficient Representation of Body Motion, Pattern Recognition Letters, vol. 31, no. 11, pp. 1369-1376, 2010.
Reference 3: N.R. Dalal and B.M. Triggs, Histograms of Oriented Gradients for Human Detection, In Proceedings of Computer Vision and Pentification Recognition (CVPR), pp. 886-893, 2005.

  In the data storage unit 104, for example, action data 301 as shown in FIG. 3 is stored as event data relating to a normal event of an object generated in advance based on video data from a surveillance camera or the like. The action data 301 stores (registers) image data of a target object extracted from video from a surveillance camera, the above-described feature amount, and label information indicating target object attribute information. As label information representing attribute information, information such as the category of the object (category such as pedestrians and bicycles), the weather at the time of shooting (sunny, cloudy, etc.), the time and time zone (daytime, evening, etc.) are stored. ing. The behavior data 301 is given a unique data ID (identification information) for each of the image data, the feature amount, and the attribute information. The registration process of the action data 301 stored in the data storage unit 104 will be described when the configuration of the data registration apparatus 300 described later is described.

  The data selection unit 105 is a feature amount similar to the feature amount of the event (for example, pedestrian behavior) of the target object created by the feature amount creation unit 103 from the behavior data 301 stored in the data storage unit 104. Collect behavioral data including As a collection method, for example, from the action data 301 in the data storage unit 104, the attribute information that matches the attribute information input for the object in the video is searched, and the action corresponding to the searched attribute information Methods such as collecting data can be used. For example, the Euclidean distance between the feature quantity created by the feature quantity creation unit 103 and the feature quantity of the action data 301 in the data storage unit 104 is calculated, and the Euclidean distance is equal to or less than a predetermined threshold. A method of collecting behavior data including feature amounts may be used. The collected behavior data is behavior data similar to the behavior of the object for which the feature amount is generated by the feature amount creation unit 103.

Further, the data selection unit 105 may collect action data from the data storage unit 104 by a data search process using an approximate nearest neighbor search method such as p-stable hashing described in Reference Document 4 below. When performing a data search process using the approximate nearest neighbor search method, the data selection unit 105 first creates a hash function using the following equation (1). Note that “a” in Expression (1) is an element value of each dimension, and the number of dimensions is the number of dimensions of the feature amount stored in the data storage unit 104. Further, “r” in the expression (1) is a width for dividing the space, and “b” is a real number uniformly selected from [0, r].

  Reference 4: M.M. Data, N.C. Immorlica, P.M. Indyk and V.M. S. Mirokni, Locality-sensitive hashing scheme based on p-stable distribution, Proceedings 20th annual Symposium on Computational Geometry, pp. 253-262, 2004.

  The data selection unit 105 creates a plurality of hash functions and configures a hash function group. FIG. 4 represents each feature quantity (referred to as feature quantity 401 in FIG. 4) of the action data 301 in the data storage unit 104 by each black circle (●) in the figure, and a feature space in which these feature quantities 401 are included. The figure divided | segmented linearly by the hash function group 402 is shown. Further, in the example of FIG. 4, it is assumed that the feature quantity 411 created by the feature quantity creation unit 103 is represented by a cross in FIG. The data selection unit 105 determines which hash function each feature quantity 401 of the action data 301 in the data storage unit 104 belongs to, and the feature quantity 411 created by the feature quantity creation unit 103. Which hash function belongs to the divided area. Further, the data selection unit 105 includes a feature in the divided region 412 to which the feature quantity 411 created by the feature quantity creation unit 103 among the feature quantities 401 of the behavior data 301 stored in the data storage unit 104 belongs. The quantity 401 (413) is specified. Then, the data selection unit 105 collects behavior data including the identified feature quantity 401 (413) from the behavior data 301 stored in the data storage unit 104. The collected action data is action data similar to the action of the object for which the feature quantity 411 is generated by the feature quantity creation unit 103.

  Next, the data selection unit 105 selects whether or not the behavior data collected from the behavior data 301 stored in the data storage unit 104 as described above is used for creating an identification model. As this selection method, for example, as shown in FIG. 5, each image 511 corresponding to each collected action data is individually combined with the video 201 and displayed on the screen of the display unit 108 to be confirmed by the user. The method of selecting with can be used. At the time of this video composition, the data selection unit 105 determines the range in which the object is moving based on the information on the input position of the region information similar to the above obtained when the action data is created by the data storage unit 104. . Then, the data selection unit 105 synthesizes the image 511 in the video 201 at different positions in accordance with the movement of each frame within the determined range. For example, the frame is switched in response to a frame switching instruction from the user via the input unit 102, and the display unit 108 displays how the image 511 moves for each frame in the video 201. The

  As shown in FIG. 5, the data selection unit 105 displays a “select” button icon 531 and a “do not select” button on the screen of the display unit 108, for example, in the lower part of the video 201 combined with the image 511. A button icon 532 is displayed. When the user inputs an instruction to the “select” button icon 531 via the input unit 102, the data selection unit 105 creates action data corresponding to the image 511 at the time of the input instruction to create an identification model. Select as learning data for the hour. On the other hand, when the user inputs an instruction to the “do not select” button icon 532, the data selection unit 105 does not select the behavior data at that time as learning data when creating the identification model. In the case of this embodiment, the selection process described above by the data selection unit 105 is repeatedly performed for each collected behavior data, and a plurality of behavior data for identification model learning is selected.

  Further, the data selection unit 105 calculates the distance between the feature amount of each collected behavior data and the feature amount from the feature amount creation unit 103, divides the behavior data according to the distance, and represents the distance for each distance. May be selected and an image of representative behavior data may be displayed on the display unit 108. In this case, an image of the representative behavior data is displayed, and the behavior data selected by the user is selected as learning data when creating the identification model.

  Then, as described above, the data selection unit 105 according to the present embodiment is configured such that the behavior data selected by the user via the input unit 102 from the behavior data collected from the data storage unit 104 is an identification model creation unit. 106.

  The identification model creation unit 106 creates an identification model using the action data selected by the data selection unit 105 as described above. As an identification model creation method, for example, a k-means clustering method may be used, and a method of creating cluster information of behavior data to obtain an identification model can be used. In this case, the number of clusters may be determined based on the number of behavior data input to the identification model creation unit 106. Then, the identification model creation unit 106 creates the centroid position and cluster range of each cluster as an identification model. In addition, when a feature amount of certain behavior data is input to the identification model, the identification model creation unit 106 determines that it is normal if the distance is within the nearest cluster in the feature space, and out of the range. If there is, an identification model for determining that it is not normal can be created. In the present embodiment, the identification model creation method is not limited to the above-described k-means clustering method, and another identification model creation method may be used.

  The identification model created by the identification model creation unit 106 is sent to and stored in the identification model storage unit 107 and also output to the display unit 108. At this time, the center of gravity of each cluster and the cluster range (for example, cluster dispersion) are output as the identification model.

The display unit 108 displays the video acquired by the video acquisition unit 101 on the screen, displays the input content by the input unit 102, and the composite image of the action data image selected by the data selection unit 105. Also display. Further, the identification model created by the identification model creation unit 106 may be displayed on the display unit 108 by an icon or the like.
The above is the configuration and processing of the information processing apparatus 100 of the present embodiment illustrated in FIG.

<Data registration processing configuration and processing>
FIG. 1B is a diagram showing a configuration part extracted from the information processing apparatus 100 shown in FIG. 1A for performing data registration processing for registering behavior data as a database in the data storage unit 104. The configuration shown in FIG. 1B may be a device different from the information processing device 100 shown in FIG. Hereinafter, in the present embodiment, the configuration shown in FIG. As shown in FIG. 1B, the data registration device 300 includes a video acquisition unit 101, an input unit 102, a feature amount creation unit 103, a display unit 108, and a data storage unit 104.

  As described above, the video acquisition unit 101 acquires video data from a surveillance camera or the like, and the video data is sent to the feature amount creation unit 103 and the display unit 108. FIG. 6 shows a screen display example of the display unit 108 on which the video 201 of the video data acquired by the video acquisition unit 101 is displayed.

  As described above, the input unit 102 acquires an input for instructing a normal event (for example, a pedestrian's behavior) of an object from the user via a GUI or the like using the screen display of the display unit 108. FIG. 6 shows an example in which an action as a normal event in which a pedestrian 621 walks on a pedestrian crossing is specified from an image 201 in which a pedestrian 621 as an object is walking on a pedestrian crossing. Further, on the screen of FIG. 6, a frame 601 is set for the pedestrian 621 in the video 201 and an attribute information list 602 is also displayed in the same manner as described with reference to FIG. When the setting of the frame 601 and the input of the attribute information using the attribute information list 602 are completed, for example, when the user gives an input instruction to the “input complete” button icon 631, the input unit 102, as described above, The area information, the attribute information, and the like are output to the feature quantity creation unit 103. On the other hand, when an input instruction to the “input complete” button icon 631 is not performed, a state in which an action can be specified and attribute information can be set is maintained.

  The feature quantity creation unit 103 creates a feature quantity in the same manner as described above. Then, the feature amount created by the feature amount creation unit 103 and the attribute information corresponding to the input by the input unit 102 are output to the data storage unit 104.

In the data storage unit 104, feature data created by the feature creation unit 103 is associated with attribute information from the input unit 102, and action data to which a data ID (identification information) is assigned is stored. Is done. As for the feature amount, for example, a hash value using the above-described p-stable hashing (information regarding which of the regions linearly divided by each hash function) is created, and these data are stored. .
In the data registration apparatus 300 of FIG. 1B, a learning database is formed by performing the data registration process as described above.

<Description of processing flowchart>
Hereinafter, the flow of processing in the information processing apparatus 100 according to the present embodiment will be described with reference to the flowcharts illustrated in FIGS. 7A to 7C. FIG. 7A shows a flowchart of an identification model creation process, FIG. 7B shows a data selection process, and FIG. 7C shows a data registration process. In the flowcharts of FIGS. 7A to 7C, steps S701 to S726 are abbreviated as S701 to S726, respectively. 7A to 7C may be executed by a hardware configuration or a software configuration, or a part of the processing may be realized by a software configuration and the rest by a hardware configuration. . When the processing is executed according to the software configuration, the processing in the flowcharts of FIGS. 7A to 7C is executed by the CPU or the like after a program stored in a ROM (not shown) is expanded in the RAM or the like. The The program according to the present embodiment is not only prepared in advance in a ROM or the like, but is also read from a removable semiconductor memory or downloaded from a network such as the Internet (not shown) and loaded into the RAM or the like. May be. The same applies to other flowcharts described later.

First, the flowchart of the identification model creation process in FIG.
In step S <b> 701, the video acquisition unit 101 acquires video data from a monitoring camera or the like, and outputs the video data to the feature amount generation unit 103 and the display unit 108. After S701, the processing of the information processing apparatus 100 proceeds to S702.
In step S <b> 702, the display unit 108 displays the video transmitted from the video acquisition unit 101. After S702, the processing of the information processing apparatus 100 proceeds to S703.

  In step S <b> 703, the input unit 102 acquires area information and attribute information as described above based on a user input instruction regarding an event of an object in the video displayed on the display unit 108, and the area information and attribute are acquired. Information is output to the feature quantity creation unit 103. After S703, the information processing apparatus 100 proceeds to S704.

  In step S <b> 704, the feature amount creation unit 103 creates a feature amount as described above based on the region information and attribute information representing the behavior of the target object, and sends the feature amount information to the data selection unit 105. Output. After S704, the processing of the information processing apparatus 100 proceeds to S705.

  In step S <b> 705, the data selection unit 105 selects action data having a feature amount similar to the feature amount created by the feature amount creation unit 103 from the database of the data storage unit 104 as described above. The detailed processing flow of the data selection processing in the data selection unit 105 will be described with reference to the flowchart of FIG. After S705, the data selection unit 105 advances the process to S706.

  In step S <b> 706, the data selection unit 105 determines whether an input completion instruction has been given from the user via the input unit 102. When the input completion instruction is not input from the user and the input regarding the action of the target object is continuously performed via the input unit 102 (NO), the data selection unit 105 returns the processing of the information processing apparatus 100 to S703. . On the other hand, when an input completion instruction is input from the user (YES), the data selection unit 105 sends the action data selected from the data storage unit 104 to the identification model creation unit 106 as data used for creation of the identification model. Output. If it is determined in S706 that an input completion instruction has been input (YES), the processing of the information processing apparatus 100 proceeds to S707.

  In S707, the identification model creation unit 106 creates an identification model (that is, learns the identification model) as described above using the action data for creating the identification model. Then, the identification model creation unit 106 stores the created identification model in the identification model storage unit 107. After completing the process in S707, the information processing apparatus 100 ends the identification model creation process.

Next, the data selection process (the process of S706) shown in the flowchart of FIG. In the following description, an example using the hash function described above will be given.
In S711, the data selection unit 105 calculates a hash value by applying a hash function as described above to the feature amount acquired in the process of S704 described above. Then, the data selection unit 105 collects behavior data having the same hash value of the feature amount stored in the data storage unit 104 with respect to the calculated hash value. If the data selection unit 105 has collected behavior data in S711, the data selection unit 105 advances the process to S712.

In S712, the data selection unit 105 initializes an index i representing a number to be assigned to the behavior data to be referred to. The index i is, for example, a number assigned in order to the action data collected as described above. When the initialization of the index i is completed, the data selection unit 105 advances the process to S713.
In step S713, the data selection unit 105 determines whether the index i of the behavior data to be referenced exceeds the number I of collected behavior data (i> I). If the index i is less than or equal to the number of collected action data (i ≦ I) (NO), the data selection unit 105 proceeds to S714, whereas if the index i exceeds the number of collected action data I (YES). Then, the process of FIG.

  In S714, the data selection unit 105 acquires the image of the image data included in the action data of the index i from the collected action data by the video acquisition unit 101 as described with reference to FIG. To the synthesized video. After S714, the data selection unit 105 advances the process to S715.

  In S <b> 715, the data selection unit 105 determines whether or not an instruction to select or not to select the action data of the index i is input via the input unit 102 by the user who has viewed the composite video displayed on the display unit 108. Determine whether. For example, when an input instruction is given to the “select” button icon 531 of FIG. 5 described above and the action data of the index i is selected (YES), the data selection unit 105 advances the process to S716. On the other hand, the data selection unit 105 advances the process to S717 when a non-selection instruction is given (NO) due to, for example, an input instruction to the “do not select” button icon 532 of FIG.

  In step S716, the data selection unit 105 sets the action data of the index i selected in step S715 as identification model creation data. After S716, the data selection unit 105 advances the process to S717.

  In step S717, the data selection unit 105 updates the index i so as to refer to the next action data, and then returns the process to step S713. Then, when the processing of S714 to S716 is completed for all the collected action data and the index i is updated in S717, the index i exceeds the number I of collected action data. Therefore, when the processing of S714 to S716 is completed for all the collected behavior data, it is determined that the number of collected behavior data I is exceeded (YES) in S713, and the processing of the flowchart of FIG.

Next, the data registration process shown in the flowchart of FIG. Note that the processing of the flowchart of FIG. 7C is performed by the data registration apparatus 300 of FIG.
In step S <b> 721, the video acquisition unit 101 acquires video data from the surveillance camera and outputs the video data to the feature amount generation unit 103 and the display unit 108. After S721, the processing of the data registration device 300 proceeds to S722.
In S722, the display unit 108 displays the video transmitted from the video acquisition unit 101. In the video in this case, the frame to be displayed is changed by a user operation via the input unit 102, and the video of the changed frame is displayed. After S722, the processing of the data registration device 300 proceeds to S723.

  In step S723, the input unit 102 acquires the above-described region information and attribute information based on a user input instruction for a normal event of the object in the video displayed on the display unit 108, and the region information and attribute are acquired. Information is output to the feature quantity creation unit 103. As described above, an action is input as a normal event by inputting a region where an object (pedestrian, bicycle, etc.) is present with respect to each frame of a video by GUI operation. After S723, the data registration device 300 proceeds to S724.

  In S724, the feature amount creation unit 103 creates a feature amount in the same manner as described above. Then, the created feature amount and attribute information are sent to the data storage unit 104. After S724, the processing of the data registration device 300 proceeds to S725.

  In step S725, the data storage unit 104 assigns and stores (registers) a data ID for the feature amount information and the registration information as described above. For the feature amount, for example, a hash value is created as described above, and these pieces of information are stored. After S725, the processing of the data registration device 300 proceeds to S726.

  In step S726, the input unit 102 determines whether an input instruction has been given to the “input complete” button icon 631 illustrated in FIG. 6 by the user. When the input completion instruction is not input from the user (NO), the input unit 102 returns the process to S723. On the other hand, when an input completion instruction is input to the input unit 102 (YES), the data registration device 300 ends the process of the flowchart of FIG. Although not shown in the example of FIG. 6, for example, when a “continue” button icon is provided on the screen and an input instruction is given to the “continue” button icon, NO is determined in S <b> 726. Then, the process may return to S723.

  As described above, in the information processing apparatus 100 according to the first embodiment, the behavior and the state of the target object in the video scene are specified, and behavior data similar to them is created and registered in advance. Collect from the database you are using. In this embodiment, appropriate action data corresponding to the video scene is selected from the action data collected from the database, and an identification model is created using the selected action data. That is, according to the information processing apparatus 100 of the present embodiment, an identification model that can accurately and accurately identify an event of an object in a video scene even if there is little video data for learning, for example, with respect to an installed monitoring camera. It is possible to create.

<Second Embodiment>
FIG. 8 shows a schematic configuration example of an information processing apparatus 800 according to the second embodiment.
When creating the identification model, the information processing apparatus 800 according to the second embodiment uses the target object as data for identifying the target object and its state in addition to the normal event of the target object described in the first embodiment. Collect behavioral data for events that differ from normal events. In the second embodiment, as an event different from the normal event of the target object, for example, when a pedestrian or bicycle is the target object, the pedestrian or the bicycle falls, falls down, or is prohibited from crossing. This includes actions such as crossing. In the second embodiment, an explanation will be given using a video scene at an outdoor intersection as an example of a video scene, but it may be a video scene of other public facilities. In the following description, an event different from a normal event is referred to as an “abnormal event”, and an action of the target object as an abnormal event is referred to as an “abnormal action”. In the case of the second embodiment, information regarding normal events and abnormal events is input by selecting and inputting a label icon described later.

  Hereinafter, in the information processing apparatus 800 according to the second embodiment illustrated in FIG. 8, input of information regarding normal and abnormal events of an object, collection of behavior data, and generation of an identification model based on the collected behavior data A description will be given of the configuration and processing to be performed. Note that in the information processing apparatus 800 of the second embodiment, the same components as those of the information processing apparatus 100 of the first embodiment described above are denoted by the same reference numerals, and description thereof is omitted. In the case of the information processing apparatus 800 according to the second embodiment, the data selection unit 805, the data storage unit 804, and the identification model creation unit 806 are different from the information processing apparatus 100 according to the first embodiment. The configuration is the same as that of the embodiment.

  In the case of the second embodiment, in addition to the input process related to the normal event of the object described in the first embodiment, the input process related to the abnormal event of the object is performed. In the input processing related to the abnormal event of the object, label information indicating the type of the event of the object and label information indicating the meaning of the abnormal event when the event of the object is an abnormal event are input. . As the label information indicating the type of event of the object, information indicating which type of the event of the object belongs to “normal” or “abnormal” is used. Further, as the label information indicating the meaning of the abnormal event of the object, for example, information such as “falling”, “falling down”, “prohibiting crossing” when the object is a pedestrian or the like is used. Therefore, in the data storage unit 804 of the second embodiment, action data in which attribute information including label information indicating the type of event of the target object and label information indicating the meaning of the abnormal event of the target object is described. Saved.

  FIG. 9 is a diagram illustrating an example of behavior data 901 when the target object is a pedestrian or the like and the behavior of the pedestrian or the like is an abnormal behavior. FIG. 9 shows behavior data 901 in which unique data IDs are assigned to the image data representing the abnormal behavior of the object, the feature amount, and the attribute information of each label information associated with the abnormal behavior. An example is given. In the attribute information of FIG. 9, “abnormality” is used as the label information indicating the type of action of the target object, “pedestrian” is used as the label information indicating the target object, and “ Examples are given of “falling”, “falling” and “no crossing”. Therefore, for example, action data 901 as shown in FIG. 9 is stored in the data storage unit 804 of the second embodiment. Although not shown in FIG. 9, information when the behavior of the object is normal is also described as appropriate in the behavior data 901.

Also in the case of the second embodiment, the data selection unit 805 selects action data used to create an identification model based on an input instruction from the user via the input unit 102.
FIG. 10 shows a display example of the display unit 108 on which the video 1001 of the video data acquired by the video acquisition unit 101 is displayed. As described above, the input unit 102 acquires an instruction input to the target object from the user via a GUI or the like using the screen display of the display unit 108. FIG. 10 shows an example of an image 1001 in which a pedestrian 1021 as an object is walking on a pedestrian crossing. In the case of the second embodiment, on the screen of the display unit 108, in addition to the video 1001 of the video data and the attribute information list 1012 similar to the above, each label information of the action data stored in the data storage unit 804 is displayed as an icon. Is also displayed. That is, the label list 1002 is a list classified based on the label information of the attribute information of each action data stored in the data storage unit 804. FIG. 10 shows an example in which a normal behavior label list corresponding to normal behavior, an abnormal behavior label list corresponding to abnormal behavior, and other behavior label lists are displayed as the label list 1002.

  In the case of the second embodiment, the user inputs label information of action data by an icon instruction in the label list 1002 by a GUI operation via the input unit 102 while viewing the video 1001 of the video data. FIG. 10 shows an example in which the user inputs, for example, a crossing prohibition label icon 1003 via the input unit 102. That is, when the pedestrian 1021 is walking on a pedestrian crossing as shown in the example of FIG. 10 and the crossing prohibition label icon 1003 is set, the behavior of the pedestrian 1021 walking on the pedestrian crossing is an abnormal behavior. The action data is set as follows.

In the present embodiment, the label list 1002 is also provided with icons that allow label information to be input to a traffic signal, for example. For example, the traffic signal 1004 for a road orthogonal to the pedestrian crossing where the pedestrian 1021 is walking is set in a red lighting state, for example, and information regarding the behavior of the pedestrian 1021 at that time is input. Is set. Thereby, it becomes possible to input action data corresponding to a change in the situation of the traffic signal 1004.
Then, when these inputs are completed, the behavior data set by these inputs is output to the identification model creation unit 806.

  The identification model creation unit 806 of the second embodiment uses the set behavior data to learn the identification model as described above. In the case of the second embodiment, each behavior data is provided with attribute information indicating the type of normal or abnormal. For this reason, the identification model creation unit 806 of the second embodiment identifies normal label information as class “+1” and abnormal label information as class “−1” using the SVM (Support Vector Machine) method. Create a model. As a result, an identification model is created that can determine whether the behavior data of the input feature quantity is normal behavior or abnormal behavior. Note that a method such as Adaboost can also be used to create the identification model.

  In the case of the second embodiment, as described above, for example, information indicating the state of a traffic signal can be input in the attribute information. For this reason, for example, an identification model for when the traffic light is lit in red is used to create an identification model for when the red light is lit. For example, an identification model for when the traffic light is lit in blue is created using the information input when the traffic light is lit in blue. It is also possible to do. In this way, by creating each identification model for lighting changes of traffic signals, for example, when identifying an object or its action, by switching the identification model based on traffic signal information, It is also possible to perform correct normal / abnormal behavior determination.

  The identification model data created by the identification model creation unit 806 as described above is sent to and stored in the identification model storage unit 107. In this embodiment, since the identification model is created using the SVM technique, a plurality of support vectors, coefficients corresponding to the respective support vectors, and threshold values are stored in the identification model storage unit 107.

Next, the identification model creation process in the second embodiment will be described in detail with reference to the flowchart of FIG. In FIG. 11, the same processing steps as those in the flowchart of FIG. 7A described above are denoted by the same reference numerals, and description thereof is omitted.
In the flowchart of FIG. 11, after S702, the processing of the information processing apparatus 800 proceeds to S1113.

  In S <b> 1113, the input unit 102 includes, on the basis of the user input instruction for the event of the target object displayed on the display unit 108, the attribute information including each piece of label information related to the abnormal behavior of the target object together with the same area information as described above Get information. After S1113, the process of the input unit 102 proceeds to S1114.

  In step S <b> 1114, the input unit 102 determines whether the label information of the attribute information instructed by the user is input from the label list 1002 described above. Specifically, the input unit 102 determines whether or not the input is performed using the label list 1002 depending on whether any label icon is selected by the user clicking a label icon in the label list 1002 or the like. . Note that the determination method is not limited to this method. If it is determined in S1114 that the input is based on the label list 1002, the instruction input information by the user via the input unit 102 is sent to the data selection unit 805, and the processing of the information processing apparatus 800 proceeds to S1115. On the other hand, if it is determined in S1114 that the input is not based on the label list 1002, the information on the instruction input by the user via the input unit 102 is sent to the feature amount creation unit 103, and the processing of the information processing apparatus 800 proceeds to S705 described above. Proceed with When the process proceeds to S704, the process of the information processing apparatus 800 proceeds to the process of S705 described above, and further proceeds to S706 described above.

  When the process proceeds to S1115, the data selection unit 805 collects behavior data from the data storage unit 804 based on the label information corresponding to the label icon input by the user via the input unit 102 in S1113. That is, the data selection unit 805 collects the behavior data searched based on the label information from the behavior data in the data storage unit 804 and sends the behavior data to the identification model creation unit 806. After S1115, the processing of the information processing apparatus 800 proceeds to S706 described above.

If it is determined in S706 that the input has been completed as described above, the processing of the information processing apparatus 800 proceeds to S1117.
In step S <b> 1117, the identification model creation unit 806 learns the identification model using the behavior data for creating the identification model. In the case of the second embodiment, the identification model creation unit 806 divides into behavior data of normal behavior and behavior data of abnormal behavior using the attribute information of the inputted behavior data. Then, as described above, the identification model creation unit 806 creates normal identification data as class “+1”, abnormal behavior data as class “−1”, and creates an identification model using SVM. The identification models (a plurality of support vectors, coefficients corresponding to the respective support vectors, and threshold values) created in this way are output to the identification model storage unit 107 and stored. After the completion of the process of S1117, the information processing apparatus 800 ends the identification model creation process of the flowchart of FIG.

  As described above, in the information processing apparatus 800 according to the second embodiment, the label icon corresponding to the label information of the action data stored in the data storage unit 804 is displayed, and the user selects the label icon. Information about the behavior of the object is input. That is, in the case of the second embodiment, it is possible to create an identification model that can determine whether an event of an object in a video scene is normal or abnormal by selecting and inputting a label icon.

<Third Embodiment>
FIG. 12 shows a schematic configuration example of an information processing apparatus 1200 according to the third embodiment.
The information processing apparatus 1200 according to the third embodiment displays the map information of the place where the video is acquired by the monitoring camera or the like together with the display of the video of the monitoring camera or the like as described in the first and second embodiments. Is also an example that allows the input of information related to the event of the object.

  Hereinafter, in the information processing apparatus 1200 shown in FIG. 12, action data as an event of an object is collected using map information including information of a plurality of scenes, and an identification model is obtained by learning based on the collected action data. The configuration and processing for creating the file will be described. In the present embodiment, an example in which surveillance cameras and the like are installed in an indoor public facility is illustrated, but the present invention is not limited to this, for example, a facility such as a hospital, a nursing facility, a station, It may be installed outdoors.

  In the information processing apparatus 1200 illustrated in FIG. 12, the map information storage unit 1201 holds the map information about the location where the surveillance camera or the like is installed and the surrounding area. The map information includes information on a floor plan (zoning map) of the building when the monitoring camera or the like is installed indoors, for example, and is stored as three-dimensional data such as CG. In addition, the map information storage unit 1201 stores monitoring camera installation information and action data of objects in each indoor area as information on a plurality of scenes with respect to information on the building. The action data for each scene can be created by CG. In addition, when the monitoring camera or the like is installed outdoors, the map information is map information around the map information. The map information in the case of outdoor includes positioning information (movement information) such as GPS (Global Radio Positioning System) mounted on a mobile phone or a vehicle, for example, as information on the behavior of the object for each scene. It may be.

  FIG. 13 shows an example in which the video 1306 acquired by the video acquisition unit 1208 of the information processing apparatus 1200 according to the third embodiment and the map 1301 supplied from the map information storage unit 1201 are displayed on the display unit 1209. Is shown. In the display example of FIG. 13, a video 1306 is a video acquired by the video acquisition unit 1208, and a map 1301 is a zoning map based on the map information supplied from the map information storage unit 1201. Further, the map information of the map information storage unit 1201 includes the installation position information of the monitoring camera and the camera information of the monitoring camera. The monitoring camera setting position information includes information on the camera installation height and camera installation angle. The camera information includes camera parameters such as the camera angle of view, focal length, aperture, shutter speed, ISO sensitivity, and pixel count. Information etc. are included. Accordingly, the monitoring camera 1302 based on the installation position information of the monitoring camera is also displayed on the map 1301 in FIG. Note that the area 1305 in the video 1306 will be described later.

  Further, in the map information of the map information storage unit 1201, data of objects of normal behavior in the area 1303 corresponding to the camera installation position information in the map 1301 is also registered. The object data included in the map information includes three-dimensional motion data of the behavior of the object. In the case of the example of FIG. 13, data of a pedestrian who moves back and forth and right and left, a person who stops, and a person who rides on a wheelchair who moves back and forth and right and left are registered as target data of normal behavior in the area 1303. , Icons 1321 to 1323 are displayed. In the map information storage unit 1201, not only normal behavior data but also target object data in the case of abnormal behavior as described in the second embodiment may be registered.

Returning to FIG.
The coordinate conversion unit 1202 reads map information, camera installation information, camera information, and data related to an object registered in the map information storage unit 1201. Then, the coordinate conversion unit 1202 performs coordinate conversion for displaying in the area of the video 1306 with respect to the area 1303 registered based on the camera installation position information and the data of the object in the area 1303. Specifically, the coordinate conversion unit 1202 calculates the area 1305 on the video 1306 by performing perspective projection conversion of the area 1303 into the video 1306 using the following equation (2) with reference to the installation position of the camera. To do.

  In equation (2), (x, y, z) are coordinates in the video 1306, k is the effective size of the pixel, o is the center (image center) of the video 1306, f is the focal length of the camera, (X, Y, Z, 1) is data of the coordinate system when the camera installation position is used as a reference. In addition, the coordinate conversion unit 1202 performs the following equation (3) to convert three-dimensional data into data in the camera coordinate system.

  In equation (3), (X, Y, Z) are coordinates in the data coordinate system, t is the camera installation position with reference to the data coordinate system, θ is the camera installation angle, and (X ′, Y ′). , Z ′) are coordinates in the camera coordinate system.

  The coordinate conversion unit 1202 can set the corresponding area 1305 in the video 1306 by performing this coordinate conversion operation on the area 1303 of the map 1301 in FIG. 13. Further, the coordinate conversion unit 1202 similarly converts a three-dimensional motion vector into a motion vector on the video 1306 for the object (icons 1321, 1322, 1323) in the area 1303 of the map 1301. Information converted by the coordinate conversion unit 1202 is output to the feature amount generation unit 1203.

  The feature amount creation unit 1203 creates a feature amount based on the motion data converted by the coordinate conversion unit 1202. Specifically, the feature amount creation unit 1203 obtains, as a feature amount, a feature vector having n frames of motion as each element of the vector from motion data on the video 1306 calculated by the conversion by the coordinate conversion unit 1202. . In addition, the feature value creation unit 1203 may create HOF feature values for n frames as feature values. Then, the feature quantity creation unit 1203 outputs the created feature quantity to the data selection unit 1205.

The data storage unit 1204 stores each action data. In the case of this embodiment, the data storage unit 1204 stores action data similar to that described with reference to FIGS.
The data selection unit 1205 uses the feature amount acquired from the feature amount creation unit 1203 to select action data having a similar feature amount from the data storage unit 104 as in the above-described embodiment. Then, the selected similar action data is sent to the identification model creation unit 1206.
The identification model creation unit 1206 creates an identification model using the action data selected by the data selection unit 1205 as in the above-described embodiment. When abnormal behavior data is also registered in the map information storage unit 1201, the identification model creation unit 1206 can create a two-class identification model such as SVM as described above. The created identification model is sent to the identification model storage unit 1207 and stored. The identification model may be sent to the display unit 1209.

A video acquisition unit 1208 according to the third embodiment acquires video data captured by a monitoring camera registered in the map information storage unit 1201. This video data is sent to the display unit 1209.
The display unit 1209 displays the video from the video acquisition unit 1208 and the identification model created by the identification model creation unit 1206. In the case of the third embodiment, on the screen of the display unit 1209, as shown in FIG. 13, icons 1321 to 1323 corresponding to the identification model are displayed overlaid on the area 1305 of the video 1306. Thereby, the user can confirm the identification result. In the example of FIG. 13, the icons 1321 to 1323 on the area 1305 are the same as those in the area 1303 of the map 1301.

Hereinafter, with reference to the flowchart of FIG. 14, processing from creation of an identification model to display in the information processing apparatus 1200 according to the third embodiment will be described in detail.
In step S1401, the coordinate conversion unit 1202 reads the above-described map information, camera installation position information, camera information, and object data registered in the map information storage unit 1201. After S1401, the process of the coordinate conversion unit 1202 proceeds to S1402.

  In step S1402, the coordinate conversion unit 1202 performs the above-described coordinate conversion processing using the map information, camera setting position information, camera information, and object data acquired from the map information storage unit 1201. Then, the coordinate conversion unit 1202 outputs the data obtained by the coordinate conversion to the feature amount creation unit 1203. After S1402, the processing of the information processing device 1200 proceeds to S1403.

In step S <b> 1403, the feature amount creation unit 1203 creates a feature amount as described above using the data converted by the coordinate conversion unit 1202, and sends the created feature amount information to the data selection unit 1205. After S1403, the process of the information processing apparatus 1200 proceeds to S1404.
In step S1404, the data selection unit 1205 selects action data of similar feature amounts as described above based on the feature amounts acquired from the feature amount creation unit 1203, and uses the selected behavior data as an identification model creation unit. 1206. After S1404, the process of the information processing apparatus 1200 proceeds to S1405.
In step S1405, as described above, the identification model creation unit 1206 creates an identification model using the selected action data, and outputs the created identification model data to the identification model storage unit 1207 and the display unit 1209. After S1405, the processing of the information processing apparatus 1200 proceeds to S1406.

  In step S <b> 1406, the display unit 1209 acquires a video from the video acquisition unit 1208 and acquires identification model information from the identification model storage unit 1207. When there are a plurality of surveillance cameras, it is assumed that which surveillance camera image is acquired in advance. After S1406, the display unit 1209 displays the video acquired from the video acquisition unit 1208 and the identification model acquired from the identification model storage unit 1207 on the screen as processing of S1407. By viewing this display, the user can confirm what identification model has been created.

  As described above, according to the information processing apparatus 1200 of the third embodiment, it is possible to automatically create an identification model that is compatible with an installed monitoring camera based on map information. Also in the information processing apparatus 1200 of the third embodiment, an identification model that can accurately identify an event of an object in a scene can be generated as in the above-described embodiment.

  The present invention supplies a program that realizes one or more functions of the above-described embodiments to a system or apparatus via a network or a storage medium, and one or more processors in a computer of the system or apparatus read and execute the program This process can be realized. It can also be realized by a circuit (for example, ASIC) that realizes one or more functions.

  The above-described embodiments are merely examples of implementation in carrying out the present invention, and the technical scope of the present invention should not be construed as being limited thereto. That is, the present invention can be implemented in various forms without departing from the technical idea or the main features thereof.

  DESCRIPTION OF SYMBOLS 101 Image | video acquisition part, 102 input part, 103 feature-value preparation part, 104 data storage part, 105 data selection part, 106 identification model preparation part, 107 identification model storage part, 108 display part

Claims (15)

Storage means for storing a plurality of event data including a feature amount of the event of the object generated in advance;
Video acquisition means for acquiring video;
A feature amount creating means for creating a feature amount of an event of an object in the acquired video;
Selecting means for selecting event data including a feature quantity similar to the feature quantity created by the feature quantity creating means from among the event data stored in the storage means;
Model creation means for creating an identification model for identifying an event of an object in a video using the feature amount of the selected event data;
An information processing apparatus comprising: Having information acquisition means for acquiring input information;
The information acquisition means acquires attribute information related to the event of the object in the acquired video, and creates event data including the feature quantity of the target in the video and the acquired attribute information The information processing apparatus according to claim 1.   The selection means is a distance between the feature quantity created by the feature quantity creation means and the feature quantity included in the event data stored in the storage means, or a predetermined approximate nearest neighbor search method The information processing apparatus according to claim 1, wherein the selection is performed using a function calculated by: The event data stored in the storage means includes attribute information regarding the event of the object,
The selecting means stores event data including attribute information that matches attribute information included in the created event data for the object in the video as event data including the similar feature amount. The information processing apparatus according to claim 2, wherein the information is selected from stored event data.   The information acquisition means acquires attribute information corresponding to a label selected from a plurality of labels representing an event of the object as the attribute information related to the event of the object in the acquired video. The information processing apparatus according to claim 2 or 4, characterized in that: Having a display means;
The information processing apparatus according to claim 1, wherein the display unit displays the video and the created identification model.   The selection means causes the display means to display an image of the object based on event data including the similar feature amount so as to be superimposed on the displayed video, and is selected from the displayed images. 7. The information processing apparatus according to claim 6, wherein event data corresponding to the selected image is selected as event data used for creating the identification model. Storage means for storing a plurality of event data including a feature amount of the event of the object generated in advance;
Information storage means for storing map information including information of a plurality of scenes;
Based on the information on the scene stored in the information storage unit, a feature amount creation unit that creates a feature amount of the event of the object;
Selecting means for selecting event data including a feature quantity similar to the feature quantity created by the feature quantity creating means from among the event data stored in the storage means;
Model creation means for creating an identification model for identifying an event of an object in a video using the feature amount of the selected event data;
An information processing apparatus comprising:   9. The information storage unit according to claim 8, wherein in addition to the scene information, the information storage unit stores information related to the event of the object and camera information related to the camera specified by the map information. Information processing device. Video acquisition means for acquiring video captured by the camera specified by the map information;
Display means,
10. The information processing according to claim 9, wherein the display unit displays the information on the scene stored in the information storage unit, the video acquired by the video acquisition unit, and the identification model. apparatus.   The said storage means preserve | saves at least any event data of the normal event of the said target object, and the abnormal event of the said target object, The any one of Claim 1 to 10 characterized by the above-mentioned. Information processing device. Model storage means for storing the created identification model;
The information according to any one of claims 1 to 11, wherein the identification model stored in the model storage unit is used to identify an event of the object in the acquired image. Processing equipment. A storing step of storing a plurality of event data including the feature amount of the event of the object generated in advance;
A video acquisition process for acquiring video;
A feature amount creating step of creating a feature amount of an event of an object in the acquired video;
A selection step of selecting event data including a feature amount similar to the feature amount created in the feature amount creation step from the event data saved in the saving step;
A model creation step of creating an identification model for identifying an event of an object in a video using the feature amount of the selected event data;
An information processing method for an information processing apparatus, comprising: A storing step of storing a plurality of event data including the feature amount of the event of the object generated in advance;
An information storage step for storing map information including information on a plurality of scenes;
Based on the information on the scene stored in the information storage step, a feature amount creation step of creating a feature amount of the event of the object;
A selection step of selecting event data including a feature amount similar to the feature amount created in the feature amount creation step from the event data saved in the saving step;
A model creation step of creating an identification model for identifying an event of an object in a video using the feature amount of the selected event data;
An information processing method for an information processing apparatus, comprising:   The program for functioning a computer as each means of the information processing apparatus of any one of Claim 1 to 12.

Images