JP2008052683A - Suspicious individual determination system and profile distribution system - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a suspicious individual determination system and profile distribution system which have high in determination precision and can perform a high-speed operation. <P>SOLUTION: Profile data which are obtained by extracting the moving speed, the moving direction, and the direction of a face of a person from a moving picture signal by a profile means 21, and a case base table stored in a database 23 in advance are compared with each other and determined by a comparison determination means 25. Profile data with a determination result stored in a case base data memory 26 added thereto are added to the database 23, and the case base table is updated. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention relates to a suspicious person discriminating system for automatically discriminating a general person and a suspicious person from a captured video signal of a camera, and a profile distribution system for distributing a case base table as data thereof.

  In recent years, surveillance cameras are often used for crime prevention and ensuring safety. Human visual monitoring requires labor costs and other operational costs, so as a system that automatically identifies suspicious persons by analyzing captured video signals, the actor is measured as a moving image and the time series of the coordinates at the center of the image An example of determining a suspicious person when the data deviates from a normal value by mathematically analyzing and quantifying data is known (see, for example, Patent Document 1).

Further, the object is continuously photographed, the velocity v and the moving direction displacement angle θ per predetermined unit time are obtained from the movement trajectory, the average value X of v · cos θ per unit time within the predetermined time is obtained, An example is known in which a predetermined extraction target is discriminated by a statistical means using an average value as an operation element index (see, for example, Patent Document 2).
JP 2005-353016 A JP 2005-275912 A

  However, the above-described conventional method or system discriminates a suspicious person using the trajectory of the center of the image of the actor, obtains the speed and displacement in the moving direction from the moving trajectory of the object, obtains the average value of v · cos θ, Since the suspicious person is discriminated by performing statistical processing, there is a problem that the discrimination accuracy of the suspicious person is not sufficient.

  In view of the above problems, an object of the present invention is to provide a suspicious person discrimination system with high discrimination accuracy by inferring a person's face orientation or usage profile as a discrimination element.

  Another object of the present invention is to provide a suspicious person discrimination system with higher discrimination accuracy by updating a case base table in consideration of discriminated profile data.

  The present invention also collects case base tables from a plurality of suspicious person discrimination systems, reconstructs the case base tables, sends them to the respective suspicious person discrimination systems, and generates the case base tables of the respective suspicious person discrimination systems. An object is to provide a suspicious person discrimination system with higher discrimination accuracy by updating.

  In order to achieve the above object, the suspicious person discrimination system according to the present invention described in claim 1 extracts profile data from a moving image signal by extracting and quantifying a person's moving speed, moving direction and face direction in time series. Profile means, a profile data memory for storing profile data generated by the profile means, a database for storing a case base table composed of a plurality of case base data including discriminated profile data and discrimination results, and the profile Comparing and determining means for determining whether or not the person is a suspicious person by comparing the profile data stored in a data memory with a case base table stored in the database. To do.

  Further, the suspicious person discriminating system according to the present invention described in claim 2 extracts a speed and direction of movement of a person from a moving image signal in time series, and quantifies it with a use as profile data, A profile data memory for storing profile data generated by the profile means, a database for storing a case base table including a plurality of case base data including the determined profile data and the determination results, and stored in the profile data memory. And comparing / determining means for determining whether or not the person is a suspicious person by comparing the profile data and the case base table stored in the database.

  The suspicious person discriminating system according to the present invention described in claim 3 stores discriminated profile data output from the comparison discriminating means and case base data including a discrimination result, and additionally stores it in the database to store the case. A case base data memory for updating the base table is further provided.

  In the suspicious person discriminating system according to the present invention as set forth in claim 4, the profile means takes values obtained by abstracting each of the moving speed, moving direction and face direction of the person using membership functions. This value is output as it is or after being time-averaged and output.

  Further, in the suspicious person discriminating system according to the present invention as set forth in claim 5, the profile means takes values obtained by abstracting the moving speed, moving direction and usage of the person using membership functions. The value is output as it is or after being time-averaged and output.

  The suspicious person discrimination system according to the present invention described in claim 6 is characterized in that the comparison discrimination means performs associative inference using the membership function.

  The profile distribution system according to the present invention described in claim 7 includes an input / output unit for inputting a part or all of the case base table from each of the plurality of suspicious person discrimination systems, and a plurality of the input / output unit from the input / output unit. Control means for inputting a case base table, reconstructing and outputting the case base table, and a database for inputting and storing the case base table from the control means, the control means storing in the database The case base table is sent to the plurality of suspicious person discrimination systems via the input / output means.

  Further, in the profile distribution system according to the present invention described in claim 8, the control means averages case base data closer than a predetermined distance for each case base data of the plurality of case base tables to form one case. Base data is used, and case base data far from the predetermined distance are directly used as the respective case base data.

  According to the suspicious person discriminating system according to the first aspect of the present invention, inference is performed by adding the profile of the human face orientation to the discriminating element, so that a suspicious person discriminating system with high discrimination accuracy can be realized.

  According to the suspicious person discriminating system according to the second aspect of the present invention, inference is performed by adding the profile of the application to the discriminating element, so that a suspicious person discriminating system with high discrimination accuracy can be realized.

  According to the suspicious person discriminating system according to the third aspect of the present invention, since the case base table is updated in consideration of the discriminated profile data, a suspicious person discriminating system with higher discrimination accuracy can be realized.

  According to the suspicious person discrimination system according to the present invention described in claim 4, a suspicious person discrimination system having high discrimination accuracy and capable of high-speed processing can be realized.

  According to the suspicious person discrimination system according to the present invention described in claim 5, it is possible to realize a suspicious person discrimination system with high discrimination accuracy and capable of high-speed processing.

  According to the suspicious person discriminating system according to the present invention described in claim 6, it is possible to realize a suspicious person discriminating system with higher discrimination accuracy and capable of high-speed processing.

  According to the profile distribution system according to the present invention as set forth in claim 7, the case base table is collected from a plurality of suspicious person discriminating systems, the profile data is reconstructed, and sent to the respective suspicious person discriminating systems. Since the case base table of the suspicious person discrimination system is updated, a suspicious person discrimination system with higher discrimination accuracy can be realized.

  According to the suspicious person discrimination system according to the present invention described in claim 8, it is possible to prevent the number of case base data from increasing infinitely.

  The best mode for carrying out the present invention will be described below in detail with reference to the accompanying drawings.

  FIG. 1 is a diagram showing a configuration of a suspicious person discrimination system according to Embodiment 1 of the present invention. The suspicious person discrimination system 2 includes profile means 21, profile data memory 22, database 23, comparison discrimination means 24, case base data memory 25, input / output means 26, and input / output terminals 27. The camera 1 supplies a moving image signal to the suspicious person discrimination system 2 of the present invention.

  The camera 1 is not a part of the suspicious person discrimination system 2 of the present invention, but is shown in FIG. 1 for convenience in order to facilitate explanation of the embodiment of the present invention.

  FIG. 2 is a diagram showing the structure of profile data stored in the profile data memory 22. The profile data 3 includes moving direction data 31, moving speed data 32, face orientation data 33, and application data 34.

  FIG. 3 is a diagram illustrating the structure of the case base data. 3A shows the structure of the case base data 6 for each case stored in the case base data memory 25. FIG. 3B shows a plurality of case base data 6 included in the database 23 in the form of a table. The structure of the example base table 7 is shown.

  As shown in FIG. 3A, the case base data 6 includes a case number 5, profile data 3, and a discrimination result 4. In the example shown in FIG. 3B, the case base table 7 includes six case base data 6, and the names of the case base data 6A to 6F are given for the sake of convenience.

  In the following description, the case number 5, the profile data 3, and the discrimination result 4 included in the case base data 6A are distinguished from others by giving the names 5A, 3A, and 4A, respectively. The same applies to 6F.

  In FIG. 1, a moving image signal is supplied from the camera 1. The moving image signal is guided to the profile means 21 and profile data 3 is generated by profiling. The generated profile data 3 is stored in the profile data memory 22.

  In the database 23, case base data 6A to 6F with the result of discriminating whether the person is a general person or a suspicious person is stored as a case base table 7 arranged in the form of a table.

  The case base table 7 included in the database 23 and the profile data 3 stored in the profile data memory 22 are supplied to the comparison discriminating means 24, and the case base data 6 with the discrimination result is generated by comparison and discrimination. It is stored in the base data memory 25.

  The case base data 6 stored in the case base data memory 25 is output from the input / output terminal 27 via the input / output means 26 and is added to the database 23 to update the case base table 7.

  The case base data 6 stored in the case base data memory 25 is output from the input / output terminal 27 via the input / output means 26.

  FIG. 4 is a suspicious person reasoning flowchart showing the functional operation of the suspicious person discrimination system 2, and a fuzzy associative reasoning composed of bidirectional associative memory (BAM) between layers (edited by Takaaki Tanibe, “Digital Signal Processing”). A suspicious person is discriminated by performing library 9 neural network and fuzzy signal processing (see pages 89 to 93 of Corona).

  The main procedure of suspicious person inference includes an information input procedure 81, a membership function calculation procedure 82, a profile data creation procedure 83, a suspicious person discrimination procedure 84, a discrimination result data output procedure 85, and a case base table update procedure 87.

  5-11 is a figure for demonstrating the detail of operation | movement of the suspicious person discrimination | determination system 2. FIG. FIG. 5 is a diagram for explaining the detection of the face orientation, FIG. 6 is a diagram for explaining the generation of the membership function according to the face orientation, FIG. 7 is a waveform diagram of the experiment for detecting the position of the person and the face orientation, and FIG. Is a diagram for explaining a membership function, FIG. 9 is a diagram for explaining a case base table, FIG. 10 is a diagram for explaining a case base inference model, and FIG. 11 is a diagram for explaining an intention recognition model.

  Hereinafter, the operation of the suspicious person discrimination system 2 will be described in detail with reference to FIGS.

  The operation of the profile means 21 will be described in the information input procedure 81 to the profile data creation procedure 83 shown in the suspicious person inference flowchart of FIG. The moving image signal supplied from the camera 1 is input in the shooting procedure 811, and the position of the person is grasped in time series in the position detection procedure 812. A method for grasping the position of a person is well known to those skilled in the art, and the description thereof is omitted (see Patent Documents 1 and 2). If there are a plurality of cameras 1, the accuracy of the three-dimensional coordinates can be improved.

  In face information detection procedure 813, the orientation of a person's face is grasped. As shown in FIG. 5A, in the present invention, the outline of the face is captured by detecting the skin color included in the photographed image, and the position of the nose and lips inside the outline is detected. The nose can be seen as a bright part at the center of the face outline, and the lips can be seen as a reddish part at the bottom of the face outline.

  In the application input procedure 814, for example, the position of an image to be photographed, such as an entrance of a building, a park, or a passage, is input. The movement speed calculation procedure 821 and the movement direction calculation procedure 822 calculate the movement speed and movement direction of the person by capturing the time passage of the position information of the person as a result of the position detection procedure 812. The face orientation calculation procedure 823 uses the detection result of the face information detection procedure 813 described with reference to FIG.

  Fig. 5 (b) shows the experimental results when the horizontal axis is time and the vertical axis is the left (down) right (up) direction and the face is consciously shaken from front to left to right to left. Yes, the movements of both the nose (dotted line) and lips (solid line) are captured, and it can be seen that the orientation of the face can be detected.

  Fig. 6 shows the experimental results when the horizontal axis is time and the vertical axis is left (down), right (up), and the person goes straight (solid line), straight → right turn (dense dotted line), straight → left turn (dashed line). It can be seen that by calculating the face orientation described with reference to FIG. 5, it is possible to detect a straight advance, right turn, and left turn of a person.

  An example of a membership function to be described later is shown superimposed on the right side of the figure.

  FIG. 7 shows an example in which the positions of a general person and a suspicious person (solid line is an x-direction component, and a dense dotted line is a y-direction component) and the face direction (dashed line) are measured by experiments.

  FIG. 7 (a) is data of a general person, and it can be seen that the person is moving fast toward the right front while slowly changing the direction of the face.

  FIG. 7B shows suspicious person data, and it can be seen that the person moves to the right front at a slower speed than the general person while swinging his / her face to the left and right.

  That is, the above experimental results indicate that it is possible to determine a suspicious person with high accuracy by taking into account the detection of the face direction. As will be described in detail later, by performing comparison discrimination between the two using fuzzy associative reasoning, it is possible to perform suspicious person discrimination with high speed and high accuracy.

  The membership function calculation procedure 824 and the profile data creation procedure 831 perform each membership function calculation using the movement speed, movement direction, and face direction of the person calculated in the above procedure as input data, and case data combining them. Profile data 3 is created as follows.

  FIG. 8 shows a method for calculating the movement direction, movement speed, and face direction of a person and examples of membership functions of the respective fuzzy rules.

  Membership functions are set for right, forward, left, and back for each person's move direction.

  Membership functions are set for stop, slow, middle, and fast for each person's move speed.

  Membership functions are set for small, medium, and large for the face direction of the person.

  An example of the membership function is shown superimposed on the experimental results on the right side of the experimental results of face orientation in FIG.

  In this case, a triangle having a width of 40 centered on 0 is a function indicating a straight line, a line bent from 0 toward the upper side of the figure is directed to the left, and a line bent from 0 to the lower side of the figure is a function indicating the right direction.

  The profile data creation procedure 831 performs the above membership function calculation procedure 824 for the moving direction, moving speed, and face direction of the person, and calculates the average value of the membership values, which are the respective calculation results, for 5 seconds. The profile data memory storing procedure 832 stores the calculation result as profile data 3 in the profile data memory 22.

  FIG. 9 is an example of the case base table 7 stored in the database 23. In this example, six cases are stored.

  In this figure, the case numbers are general persons 1 to 3 and suspicious persons 1 to 3, and profile data 3 is for the movement direction of the person, the movement speed of the person, and the face direction of the person. In addition, it is shown whether the result calculated by the information input procedure 81 and the membership function calculation procedure 82 described above is a normal person or a suspicious person (abnormal).

  The suspicious person discrimination procedure 84 is performed using the profile data 3 and the case base table 7 described with reference to FIG.

  The profile data reading procedure 841 reads the profile data 3, the case base table reading procedure 842 reads the case base table 7, and the suspicious person discrimination procedure 843 performs discrimination.

  FIG. 10 is a diagram for explaining a case-based reasoning model. The suspicious person discrimination procedure 843 can perform suspicious person discrimination with high accuracy and high speed by performing fuzzy associative reasoning as shown in the case-based reasoning model. In the figure, the lowest level (Input Layer) corresponds to profile data 3, and the middle level (Case Base Layer) corresponds to case base table 7. A large number of diagonal lines in the figure conceptually indicate that inference is performed with reference to the respective membership values. The inference results are collected in the upper stage (Output Layer), and it is determined whether the profile data 3 is a normal person or a suspicious person (abnormal).

  As shown in the intention recognition model of FIG. 11, the suspicious person determination procedure 843 can perform suspicious person determination with high accuracy and high speed by performing fuzzy associative inference in addition to the inference element.

  In the figure, the bottom level (Input Layer) corresponds to profile data 3, the middle level (Case Base Layer) corresponds to case base table 7, and the top level (Output Layer) corresponds to discrimination result 4 in the case base reasoning explained in FIG. Similar to the model, but the position where the camera 1 is installed is, for example, a passage, a park, or the entrance of a building is regarded as a human intention for each purpose, and a membership function is defined using this to calculate a membership function. It is possible to generate the profile data 3 by performing the procedure 824. Also in this case, the inference results are collected in the upper stage (Output Layer), and it is determined whether the profile data 3 is a normal person or a suspicious person (abnormal).

  In the determination result data output procedure 85, the case base data 6 obtained by adding the case number 5 and the determination result 4 to the profile data 3 is stored in the case base data memory 25 and is output to the input / output terminal 27 by the input / output means 26.

  The registration determination procedure 86 determines whether or not to add the case base data 6 stored in the case base data memory 25 to the case base table 7. If NO, the process proceeds to the end determination procedure 88. If this is the case, this is updated in the case base table 7 and the process proceeds to the end determination procedure 88.

  Whether to add the case base data 6 to the case base table 7 can be determined automatically by calculating the certainty of the determination result and referring to this, but can also be performed manually. .

  As described above in detail, the suspicious person discrimination system according to the first embodiment of the present invention can perform suspicious person discrimination with high accuracy and high speed.

  In addition, this invention is not limited to the said Example.

  In the above description using the suspicious person inference flowchart shown in FIG. 4, the profile data memory storing procedure 832 stores the average value of the membership values for 5 seconds. However, the time average need not be limited to 5 seconds, and the camera Needless to say, the time average value may vary depending on the installation status of 1 and the content of the video signal.

  Also, the steps of the suspicious person inference flowchart shown in FIG. 4 do not necessarily have to be in this order. For example, the position detection procedure 812 and the face information detection procedure 813 may affect the processing result even if they are in succession. Needless to say, the order of the procedures not given may be changed as necessary.

  Further, in the suspicious person reasoning flowchart shown in FIG. 4, an example in which suspicious person determination is performed by performing fuzzy associative reasoning has been described. However, another determination method such as a combination of simple threshold determinations may be used. Needless to say.

  FIG. 12 is a diagram showing a configuration of a profile distribution system according to the second embodiment of the present invention. In the figure, 1A to 1C are cameras, 2A to 2C are suspicious person discrimination systems, 27A to 27C are input / output terminals, and 70 is a server device.

  The server device 70 includes an input / output unit 71, a control unit 72, and a database 73. The input / output means 71 is connected to the input / output terminals 27A to 27C, and the contents of the respective case base tables 7 are introduced into the database 73 from the suspicious person discrimination systems 2A to 2C according to instructions from the control means 72, The included case base table 731 can be transmitted as the case base table 7 of the suspicious person discrimination systems 2A to 2C.

  FIG. 13 is a flowchart for explaining the operation of the server device 70.

  Hereinafter, the operation of the profile distribution system according to the second embodiment of the present invention will be described in detail. The combination of the camera 1A and the suspicious person discrimination system 2A is the same as the suspicious person discrimination system according to the first embodiment of the present invention, and the suspicious person discrimination by the combination of the camera 1B and the suspicious person discrimination system 2B and the camera 1C and the suspicious person discrimination system 2C. Since the system is the same, description of the operation is omitted.

  The example shown in FIG. 12 is an example of three suspicious person discrimination systems 2 of 2A to 2C, but the same operation is possible if there are a plurality of suspicious person discrimination systems 2.

  In the data collection procedure of FIG. 13, the control means 72 introduces the contents of each case base table 7 from the suspicious person discrimination systems 2 </ b> A to 2 </ b> C into the database 73.

  The case base table 731 is reconstructed by integrating the plurality of case base tables 7 collected above in the reconfiguration procedure 91. At this time, the case base data whose membership function values are closer than a predetermined distance average the membership functions of the two into one data. As a result, the number of case base data can be prevented from increasing indefinitely.

  In the data redistribution procedure 92, the case base table 731 is transmitted as the case base table 7 of the suspicious person discrimination systems 2A to 2C. Since more case base data can be collected, the discrimination accuracy can be improved.

  The timing at which the data collection procedure 90 starts operating may be determined manually or periodically according to a request from at least one of the suspicious person discrimination systems 2A to 2C connected to the input / output terminals 27A to 27C. The same applies to the timing at which the data redistribution procedure 92 starts operation.

  As described above in detail, by using the profile distribution system according to the second embodiment of the present invention, it is possible to perform suspicious person discrimination with higher accuracy and speed.

  In addition, this invention is not limited to the said Example.

  Although the server device 70 is provided separately from the suspicious person discrimination systems 2A to 2C, it goes without saying that one or a plurality of the suspicious person discrimination systems 2A to 2C may have the function of the server device 70.

It is a figure which shows the structure of the suspicious person discrimination | determination system by Example 1 of this invention. It is a figure which shows the structure of the profile data memorize | stored in the profile data memory of Example 1 of this invention. It is a figure which shows the structure of the example base data of Example 1 of this invention. It is a suspicious person inference flowchart which shows the functional operation | movement of the suspicious person discrimination | determination system by Example 1 of this invention. It is a figure explaining the detection of the direction of a face. It is a figure explaining the production | generation of the membership function by the direction of a face. It is a wave form diagram of an experiment which detects a person's position and face direction. It is a figure explaining a membership function. It is a figure explaining an example base table. It is a figure explaining the case-based reasoning model for demonstrating operation | movement of the comparison discrimination | determination means of the suspicious person discrimination | determination system by Example 1 of this invention. It is a figure explaining the intention recognition model for demonstrating operation | movement of the comparison discrimination | determination means of the suspicious person discrimination | determination system by Example 1 of this invention. It is a figure which shows the structure of the profile distribution system by Example 2 of this invention. It is a flowchart for demonstrating operation | movement of the profile distribution system by Example 2 of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Camera 2 Suspicious person discrimination | determination system 21 Profile means 22 Profile data memory 23 Database 24 Comparison discrimination means 25 Case base data memory 26 Input / output means 3 Profile data 70 Server apparatus 71 Input / output means 72 Control means 73 Database

Claims (8)

Profile means that extracts the speed, direction of movement, and face direction of the person from the video signal in time series and quantifies them as profile data;
A profile data memory for storing profile data generated by the profile means;
A database for storing a case base table including a plurality of case base data, including determined profile data and determination results;
Comparing and determining means for determining whether or not the person is a suspicious person by comparing the profile data stored in the profile data memory and a case base table stored in the database. Suspicious person discrimination system characterized. Profile means that extracts the speed and direction of movement of the person from the video signal in time series, and quantifies it with the application as profile data;
A profile data memory for storing profile data generated by the profile means;
A database for storing a case base table including a plurality of case base data, including determined profile data and determination results;
Comparing and determining means for determining whether or not the person is a suspicious person by comparing the profile data stored in the profile data memory and a case base table stored in the database. Suspicious person discrimination system characterized.   A case base data memory that stores discriminated profile data and discrimination results output from the comparison discriminating means, and additionally stores in the database and updates the case base table is further provided. The suspicious person discrimination system according to claim 1 or 2.   The profile means uses values obtained by abstracting the moving speed of the person, the direction of movement, and the direction of the face using membership functions, and outputs the values as they are or after performing time averaging processing. The suspicious person discrimination system according to claim 1 characterized by things.   The profile means uses values obtained by abstracting each of the moving speed, moving direction, and usage of the person using membership functions, and outputs the values as they are or as they are time-averaged and output. The suspicious person discrimination system according to claim 2.   6. The suspicious person discrimination system according to claim 4 or 5, wherein the comparison discrimination means performs associative inference using the membership function. Input / output means for inputting a part or all of the case base table from each of the plurality of suspicious person discrimination systems according to any one of claims 1 to 6;
Control means for inputting a plurality of the case base tables from the input / output means, reconstructing them into one case base table, and outputting them;
A database for inputting and storing a case base table from the control means, and the control means stores the case base table stored in the database to the plurality of suspicious person discrimination systems via the input / output means. A profile distribution system characterized by sending. For each case base data of the plurality of case base tables, the control means averages case base data closer than a predetermined distance into one case base data, and case base data farther than the predetermined distance remain as they are. The profile distribution system according to claim 7, wherein each case base data is used.