JP2012208793A - Security system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a security system which is available in a versatile manner by defining and creating a flexible determination condition for detection in accordance with a destination of installation or a purpose of use when detecting suspicious persons.SOLUTION: A security system comprises a feature data DB 31 storing therein feature data associated with each person, a determination file 32 storing therein a determination condition defined from a combination of feature keywords, a determination unit 40 detecting, on the basis of matching comparison between a feature data value of feature data and the determination condition in the determination file, a person associated with the feature data as a detection person, and an output unit 50 which outputs a control signal to an output destination when the detection person is detected. The determination file is comprised of the determination condition defined by combining different types of a plurality of feature keywords using a connecting key, the output destination to output person detection when it is determined that the determination condition is satisfied, and contents of the control signal to the output destination.

Description

  The present invention relates to the field of security devices that perform person detection having predetermined characteristics.

  In recent years, security systems and security devices have been widely installed due to an increase in crime prevention awareness. Heretofore, a security device that detects a suspicious person based on image analysis processing of an image captured by a surveillance camera or the like has been known.

  For example, Patent Document 1 describes an alarm device that determines that there is a possibility of a suspicious person when a person whose face is hidden by sunglasses or a mask is detected from an image, for example, at an ATM or entrance of a financial institution. Has been. More specifically, when the image processing unit calculates the face concealment score as to whether the eyes and mouth of the face have been reliably detected, using a unique calculation formula, and the determination unit determines that the score value is greater than a threshold value that is suspicious Is determined as a suspicious person whose face is hidden by sunglasses or a mask.

Japanese Patent No. 2648054

  Here, even if a suspicious person is generally detected, the purpose of use differs depending on the installation location (use location) of the security device. In other words, what kind of characteristics a person is considered as a suspicious person who wants to detect differs depending on the needs of the side using the security device.

  However, in conventional security devices that detect suspicious persons, the characteristics and behavior of suspicious persons are defined according to the installation location and purpose of use, and the detection is specialized for the required image processing and the intended use of the installation location. Since the determination method is implemented, there is a problem that when the same apparatus or system is used for another installation destination or for another purpose, it is necessary to re-install a detection determination program.

  For example, in hospitals and clinics, there are many people wearing only a mask due to a cold or the like, and these people should not be detected as suspicious people. For this reason, what is required in hospitals and the like is a security device that detects a person wearing both sunglasses and a mask and a person wearing only sunglasses as a suspicious person.

  However, although the security device described in Patent Document 1 can detect a person wearing both sunglasses and a mask, a person wearing both sunglasses and a mask, and a person wearing only sunglasses. A person who is detected as a suspicious person and wears only a mask cannot make a detection decision that meets the needs of a hospital or the like not to detect it. In the first place, the security device described in Patent Document 1 is a device implemented assuming installation and use in a financial institution, etc., and a person who hides most of his face with sunglasses or a mask is considered a suspicious person. This is because it is detected. And if you want to make such a detection decision, detect people wearing sunglasses and a person wearing masks, and do not detect people who wear only masks against people. It is necessary to change the judgment conditions and implement the program. That is, in the conventional apparatus for detecting a suspicious person, if the installation destination and purpose of use change, it will be necessary to re-install the detection judgment program. It is not easy to use with.

  In view of the above points, the present invention defines and creates flexible detection criteria according to the installation destination and purpose of use when detecting a suspicious person, regardless of the installation destination or purpose of use. It aims at providing the security device which can be used universally.

  In order to solve the above-described problem, the security device according to the present invention includes a first storage unit that stores feature data including a plurality of types of feature data values indicating the characteristics of a person in association with each person, and the feature data value. A second storage unit that stores a determination file including a determination condition defined from a combination of feature keywords corresponding to the feature of the person indicated by the character string, and a match comparison between the feature data value of the feature data and the determination condition of the determination file A determination unit that determines whether or not the determination condition is satisfied, and when the determination condition is satisfied, a determination unit that detects a person associated with the feature data as a detected person, and an output when the detected person is detected An output means for outputting a control signal for the destination, and the determination file includes a determination condition defined by combining a plurality of different types of feature keywords using a connection key. When, characterized in that it is configured with the output destination to output the person detected when it is determined that the determination condition is satisfied, by the content of the control signal to the output destination.

  In order to solve the above problem, the security device provides a screen for creating the determination file, and one feature representation is selected from the feature representations indicating the characteristics of the person to be detected on the screen. A combination of a plurality of different types of feature keywords using the feature keyword or the connection key previously associated with the feature expression, and defining from the created feature keyword or the combination of feature keywords And a determination file creating means for creating a determination file including the determination conditions to be determined, and the determination file created by the determination file creating means is stored in the second storage means.

  In order to solve the above-mentioned problem, in the security device, the determination means sets a value of “true” or “false” for each feature keyword based on a match comparison with a feature data value corresponding to the feature keyword. Output, and then, when the output “true” or “false” value is combined with the connection key, it is determined whether or not the determination condition is satisfied. If not satisfied, a “false” value is output.

  In order to solve the above-described problem, in the security device, the determination unit is a case where a predetermined function is included in the determination condition, and is designated by a first parameter of the predetermined function for each feature keyword. When outputting a value of “true” or “false” based on comparison with the feature data value corresponding to the feature keyword from the feature data for the designated number, the feature data for the designated number When the specified number of feature data specified by the second parameter of the predetermined function matches the feature keyword, a value of “true” is output.

In order to solve the above-mentioned problem, in the security device, at least input means for inputting an image from a camera, image analysis processing is performed on the image, and person characteristic data analyzed in the image is stored for each person. The feature data aggregated for each person by the image analysis means is stored in the first storage means.

In addition, what applied the arbitrary combination of the component of this invention, expression, or a component to a method, an apparatus, a system, a computer program, a recording medium, etc. is also effective as an aspect of this invention.

  According to the present invention, when detecting a suspicious person, by defining and creating a flexible determination criterion for detection according to the installation destination and purpose of use, it can be used regardless of the installation destination or purpose of use. A security device that can be used can be provided.

The block diagram of one Example which shows the whole structure of the security system 100 which concerns on this embodiment is shown. It is a figure explaining the relationship between feature data and a determination file. It is an example of feature data. A feature data specification table is shown. It is an example (the 1) of the determination file 32. FIG. An example sentence of a judgment condition sentence is shown. The feature keyword specification table is shown. A function specification table is shown. A connection key specification table is shown. It is an example (the 2) of the determination file 32. 6 is an example of feature items and feature data values extracted from a plurality of feature data. It is a flowchart explaining a feature data generation process. It is a flowchart explaining the judgment process of a suspicious person. It is a flowchart explaining the output process of a determination result. An example (part 1) of a maintenance application setting screen is shown. An example (part 2) of a maintenance application setting screen is shown.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention will be described with reference to the drawings in each embodiment.

[Overview]
In the security device according to the present embodiment, features and actions that are desired to be detected (detected) as a suspicious person, and their locations and times are defined in detail. Finally, these features and actions are defined as and, or, or next, etc. Prepared and implemented a general method for creating judgment files that can flexibly create detection criteria defined in relations. Then, the judgment file created by this method is compared with the characteristics and behaviors based on the actual image processing result, so that the suspicious person is detected and judged.

  By doing so, in the security device according to the present embodiment, when detecting a suspicious person, it is possible to define a flexible determination condition according to the installation destination and purpose of use, that is, to create a determination file. It can be used in a general-purpose environment regardless of the destination or purpose of use. In addition, when a new crime or the like occurs after the installation, the security service corresponding to the crime or the like can be easily started by redefining the determination condition corresponding to the crime or the like. Details will be described below.

[System configuration]
FIG. 1 is a block diagram of an example showing an overall configuration of a security system 100 according to the present embodiment. The security system 100 according to the present embodiment includes a security device 1, a data collection device 2 including a camera 2a, a sensor 2b, a monitoring center 3, and an alarm device 4.

  The security device 1 includes an input unit 10, an image analysis processing unit 20, a storage unit 30, a determination unit 40, an output unit 50, and a determination file creation unit 60.

  The input unit 10 is an input function unit that receives input data such as images and detection signals from the data collection device 2 including the camera 2a and the sensor 2b. That is, the input unit 10 takes in data from an external device into the security device 1.

  The image analysis processing unit 20 performs data analysis of input data such as images and detection signals, extracts information (feature data) necessary for detecting a suspicious person, and shapes the data in a form that can be easily used in subsequent determination processing. Or gather together.

  The storage unit 30 stores a feature data DB 31 in which character feature data is stored in association with each person, and a determination file 32 in which a feature to be detected as a suspicious person is defined as a determination condition based on the character feature data. It is a storage means. Specific examples of the feature data DB 31 and the determination file 32 will be described later.

  Based on the determination file 32, the determination unit 40 compares and collates the analyzed characteristic data of the person, and determines whether the person should be detected as a suspicious person. When it is determined that the person is to be detected as a suspicious person, the output unit 50 is instructed to output a warning to the monitoring center 3 and the alarm device 4.

  The output unit 50 is connected to the monitoring center 3 and the alarm device 4, and performs notification to the monitoring center 3 and output to the alarm device 4.

  The determination file creation unit 60 creates the determination file 32. In the security device 1 according to the present embodiment, the user can flexibly define a characteristic to be detected as a suspicious person as a determination condition depending on the installation location of the security device 1, the purpose of use, and the like. Details of this creation method will be described later again.

  The camera 2a is a surveillance camera that is provided at various locations within an installation location (such as a financial institution, a hospital, or a store) where security is performed. It is provided at an entrance / exit, a room, or a position that becomes a blind spot, and takes a video, that is, a continuous still image in real time, and inputs the captured still image to the security device 1. In view of recent camera performance, for example, about 20 to 30 images (frames) are captured per second and input to the security device 1.

  The sensor 2b is a detection sensor provided everywhere in the installation location where security is performed. Specifically, a sensor such as a magnet switch that detects opening / closing of an entrance door and outputs a detection signal, for example, a sensor such as an infrared detection sensor that detects the presence of a person installed in a hallway or a room and outputs a detection signal is there.

  The monitoring center 3 and the alarm device 4 are alarm output destinations when an abnormality occurs. The monitoring center 3 is connected to the monitoring center 3 via a communication line such as a telephone line. When the monitoring center 3 receives an alarm from the security device 1, the monitoring center 3 dispatches a security staff to the site or contacts the police, for example. The alarm device 4 is an alert lamp, an alert alarm, or the like. When an alarm is received from the security device 1, the alarm device 4 turns on the lamp or sounds an emergency alert.

  Needless to say, each function of the security device 1 described above is actually realized by a computer by a program executed by the CPU of the device. In addition, in the present embodiment, it is only one configuration example, and each function can be mounted on another device. For example, the storage unit 30 of the security device 1 can be constructed by a separate data storage server. The determination file creation unit 60 may be mounted on a terminal PC or the like, and configured to transfer the determination file created by the determination file creation unit 60 of the terminal PC to the storage unit 30 of the security device 1.

[Feature data and judgment file]
FIG. 2 is a diagram for explaining the relationship between the feature data and the determination file.

  The security device 1 performs image analysis processing every time an image (frame) and a detection signal are input, and from the result of the image analysis processing, feature data for each person ID (given for each person using a face authentication result or the like) Are generated and stored in the feature DB 31 as needed. The determination file 32 is a file in which a characteristic to be detected as a suspicious person based on person characteristic data is defined as a determination condition. The determination file 32 is created in advance by the user or via the determination file creation unit 60 and is held in the storage unit 30. The

  Then, as shown in the figure, the determination unit 40 compares and collates the determination file 32 with the feature data for each person, and determines whether or not the person is to be detected as a suspicious person. When it is determined that the person is to be detected as a suspicious person, the output unit 50 is instructed to output a warning to the monitoring center 3 and the alarm device 4. This will be described in detail below.

(Feature data)
As described above, the image input from the camera 2a and the sensor 2b and the image analysis result obtained by analyzing the detection signal are accumulated in the feature DB 31 as feature data. That is, the security device 1 performs an image analysis process every time an image and a detection signal are input, and based on the result of the image analysis process, feature data for each person ID (given for each person using a face authentication result or the like) Generate and store and update the feature DB 31 as needed.

  FIG. 3 is an example of feature data. In addition, FIG. 4 shows a specification table of feature data. The feature data in FIG. 3 shows one actual feature data. That is, when a person is detected by one image input from the camera 2a and the detection signal input from the sensor 2b, the characteristics and behavior of the person are analyzed by image analysis according to each item of the specification table of FIG. A feature data value for each feature data item is detected. Then, one feature data is generated for one image (one frame) and one person. On the other hand, if no person is detected, the feature data need not be generated.

  The specification table of FIG. 4 is an example of features and actions that can be analyzed by an image input from the camera 2a and a detection signal input from the sensor 2b. As long as an image analysis method capable of analyzing / extracting the relevant feature and action is implemented in the security device 1, the values of these feature and action items can be obtained. In addition, the detection signal input from the sensor 2b is used for specifying the presence / position of a person. Since it is possible to analyze the presence / position of a person only from an image, it is used only as supplementary information for image analysis.

  Specifically, refer to FIG. For example, the feature data item “Date” (date) indicates the date on which the feature data generation source image or detection signal was input. The feature data item “Time” (time) indicates the time when the feature data generation source image or detection signal is input.

  Also, the feature data item “Gender” (gender) indicates the gender value of the person analyzed by image analysis when one person (person ID: 0001) is captured in one input image. Here, since the gender of this person was analyzed as “male”, the gender value is “0”. “Sunglasses” (presence / absence of sunglasses) indicates the presence / absence value of the person's sunglasses analyzed by image analysis. Here, since the person's sunglasses presence / absence value is analyzed as “Yes”, the value of sunglasses presence / absence is “1”.

  “Stay [x]” (in area x) indicates the position (area) value of the person analyzed by the image analysis and the detection signal. Here, since the position (area) value of this person has been analyzed as “in area 1”, the value of Stay [1] is “1”. The values of Stay [2] and Stay [3] are “0”. “Rove” indicates whether or not the person analyzed by the image analysis is missing. Here, this value is “0” because it has been analyzed that this person is not ready.

  The feature data is generated by image analysis for each frame (captured image). A person ID is issued for each person, and is stored and managed in the feature DB until the number of frames that can be accumulated is exceeded. The determination of whether or not they are the same person may be performed using a conventional face authentication processing technique or the like. When a plurality of persons are captured in one input image, the characteristics and behaviors of each person in the image are subjected to image analysis, and feature data for the number of persons is generated.

(Judgment file)
FIG. 5 is an example (part 1) of the determination file 32. As described above, the determination file 32 is a file in which a feature to be detected as a suspicious person based on the person feature data is defined as a determination condition, and is created in advance by the user and held in the storage unit 30. Specifically, as shown in the figure, the determination file is made up of each element of “controlling IO (Input / Output)”, “control contents”, and “determination condition statement”.

  “IO to be controlled” indicates an ID (1, 2,...) Of an output target (for example, the monitoring center 3 and the alarm device 4) to be controlled when the condition by the “judgment condition sentence” is satisfied. “Control content” indicates the content of control for an output target to be controlled when the condition by the “judgment condition statement” is satisfied. “Open” means ON, “Close” means OFF. If the output target is the monitoring center 3, the notification is turned ON / OFF, and if the alarm device 4, the lamp is turned on / off, or the alarm is turned ON / OFF. I do. Since the control contents for the output target are usually “open” (ON) and “closed” (OFF), the determination conditions for “open” control and “closed” control should be defined as a set. To do. The “judgment condition sentence” is a judgment condition syntax in which a feature to be detected as a suspicious person is defined based on person feature data. This point will be described below.

  FIG. 6 shows an example of a judgment condition sentence. As shown in the figure, the configuration of the “judgment condition sentence” is roughly divided into three parts a to c. First, in the case of “Sunglasses”, for example, “Sunglasses”, “true” is output when sunglasses are detected in the feature data item of the feature data image-analyzed from the image. Outputs “false” when it is not detected. For example, in the case of “Mask”, “true” is output when the presence of a mask is detected in the feature item of the feature data, and “false” is output when the presence of a mask is not detected. Therefore, it can be seen from this that the “feature keyword” and the feature item of the feature data (FIG. 4) are in a correspondence relationship.

  FIG. 7 shows a specification table of feature keywords. The feature keywords include not only “Sunglasses” and “Mask”, but also a plurality of feature keywords as shown in the figure. Therefore, the user uses the feature keyword that the user wants to use from the specification table, that is, uses the feature keyword that should be detected as a suspicious person to define the determination condition.

  Referring again to FIG. 6, if the “individual condition” is an individual condition defined using the function FrameCount, for example, “FrameCount (x, y, z)”, Determine whether the number of times the feature keyword x corresponding to the feature data item (feature data value) of the feature data image-analyzed from the image of the most recent specified frame is “true” is at least the specified number y It is a condition. For example, in the case of “FrameCount (Sunglasses, 2,10)”, it corresponds to the feature data item (feature data value) of the feature data image-analyzed from the image of the latest 10 frames out of the latest specified 10 frames. This is a condition for determining whether or not the number of times that the characteristic keyword “Sunglasses” is “true” is at least twice.

  When a video is captured by the camera 2a, images from the camera 2a are continuously input to the security device 1. When the person is moving, the feature data is generated in units of frames (captured images), so that the feature data is generated one after another whenever the person image from the camera 2a is continuously input to the security device 1. It will be. However, in the image analysis, although the detection of the person detection itself can be detected with high accuracy, sometimes the fine feature data value of the person cannot be accurately analyzed. For example, a person wearing sunglasses cannot be analyzed from an image of the person facing back. Moreover, even if it is the image which imaged the person from the front and the side, it cannot necessarily analyze sunglasses wearing by analysis. Therefore, the feature keyword “Sunglasses” is used to determine whether or not sunglasses are worn based on the feature data value (Sunglasses = 0or1). Since it is necessary to absorb the variation, a function for improving the accuracy of feature detection is interposed in the feature keyword here. As a result, when a certain number of frames (for example, 10 frames) is provided, and the number of frames specified by the user (for example, twice) can be detected, the “individual condition” as a whole is “true”. "Is output.

FIG. 8 shows a function specification table. The output contents of each function are as shown in the figure, but the types of functions are broadly prepared for the Frame system and the Time system. For example, FrameCount is a function that outputs "true" when the feature data corresponding to the feature keyword outputs more than the number of detected frames, and outputs "false" otherwise. It is. On the other hand, TimeCount is set to “true” when the total of the time when the feature data corresponding to the feature keyword output “true” exceeds the detection time in the latest target time, and “false” otherwise. Function to output.
In other words, from the viewpoints of frame and time, it is possible to detect a person having the relevant characteristics while absorbing the above-described variations.

  Referring to FIG. 6 again, the “total condition” (judgment condition sentence) of c is connected to the “individual condition” of a with connection keys (connection words) such as and, or, next, etc. It consists of individual conditions. For example, “FrameCount (Sunglasses, 2,10)”, which is the “individual condition”, outputs “true” when a wearer of sunglasses is detected twice in 10 frames, and “FrameCount (Mask, 1,10 “)” Outputs “true” once in 10 frames when a mask wearer is detected. And in the “total condition” in which these two individual conditions are connected by “and”, “true” is set when sunglasses wearing and mask wearing are detected, that is, when both of the individual conditions are both “true”. Output.

  FIG. 9 shows a connection key specification table. Since the output contents of each connection key number are as shown in the figure, the detailed description is omitted. As described above, the user can perform a detection determination of a suspicious person suitable for the installation destination and the purpose of use of the security device 1 by creating a determination condition sentence by freely combining a to c.

[Judgment example]
An example of determining whether or not a person (for example, person ID: 0001) is a suspicious person in the determination condition of the determination file 32 based on FIGS.

  FIG. 10 is an example (part 2) of the determination file 32. As shown in the figure, the determination file 32 defines at least two sets of “determination condition statements” for “controlling IO” IDs 1 and 2.

  FIG. 11 is an example of feature items and feature data values extracted from a plurality of feature data. That is, from the feature data for the last 10 frames in which the same person (for example, person ID: 0001) is imaged, the feature items “Sunglasses” (Sunglasses), “I am in area 1” (Stay [1]) , “I am in area 2” (Stay [2]) was extracted and the value (0 or 1) was shown in the graph. In addition, feature items “Rove” were extracted from the feature data of the latest frame within 10,000 msec in which the same person was imaged, and the value (0 or 1) is shown in the graph.

  First, a determination is made regarding “FrameCount (Sunglasses, 2, 5)” which is an “individual condition” where “IO to be controlled” is ID1 and “control content” is open. This individual condition means a condition for determining whether or not the number of times that the feature keyword “Sunglasses” is “true” (Sunglasses = 1 in the feature data) among the most recent five frames is two or more. is there. From FIG. 11, sunglasses are present in 1, 2, 3, 5 frames among the latest 5 frames. Accordingly, the determination unit 40 outputs “true” to “FrameCount (Sunglasses, 2, 5)”.

  Next, a determination is made regarding “FrameCount (Stay [1], 3, 5)”, which is an “individual condition” where “IO to be controlled” is ID1, and “control content” is open. The meaning of this individual condition is whether the number of times the characteristic keyword “Stay [1]” is “true” (Stay [1] = 1) out of the last five frames is three or more. It is a condition to do. From FIG. 11, in the last five frames, it is in area 1 in 1, 2, 3, 5 frames. Therefore, the determination unit 40 outputs “true” to “FrameCount (Stay [1], 3, 5)”.

  Finally, the “total condition” (judgment condition statement) here is “FrameCount (Sunglasses, 2,5)” and “FrameCount (Stay [1], 3,5)”, which are “individual conditions”. Connected with the and connection key. As described above, according to the determination result of each “individual condition”, “true” and “true” are obtained, so that the determination unit 40 finally determines “FrameCount (Sunglasses, 2, 5) and “FrameCount (Stay [1], 3,5)” are output as “true”. As a result, “IO to be controlled” outputs “ON” as “control content” to the output target of ID1.

  On the other hand, a determination is made regarding “Not (FrameCount (Stay [1], 3, 5))”, which is an “individual condition” where “IO to be controlled” is ID1 and “control content” is closed. This individual condition means that the number of times that the characteristic keyword “Stay [1]” is “true” (Stay [1] = 1) out of the five most recent frames is not three. It is a condition. From FIG. 11, in the last five frames, it is in area 1 in 1, 2, 3, 5 frames. Therefore, the determination unit 40 outputs “false” to “Not (FrameCount (Stay [1], 3,5))”. Since there is only one “individual condition” here, this corresponds to the “total condition”. The determination unit 40 finally outputs “false” to “Not (FrameCount (Stay [1], 3,5))” which is the “total condition”. As a result, “control IO” does not output “closed (OFF)” that is “control content” to the output destination of ID1.

  Based on the above, the determination condition sentence with “controlling IO” being ID1 and the determination result for the output target destination with “controlling IO” being ID1 will be considered. That is, according to the determination condition sentence, when the person wearing sunglasses is in Area 1 (for example, a no-go-in area), the determination file 32 has a purpose of using the ID1 output target (for example, the alarm device 4). Is defined and created. In the actual feature data, when a feature or action person satisfying this determination condition is detected, the security device 1 outputs open (ON) to the alarm device 4, and the alarm device 4 sounds an alert. It is.

  After that, if this person moves out of area 1 and does not satisfy “FrameCount (Stay [1], 3,5)”, conversely, “FrameCount (Sunglasses, 2,5) and FrameCount (Stay [ “1”, 3,5) ”will be output as“ false ”, and“ Not (FrameCount (Stay [1], 3,5)) ”will be output as“ true ”. The device 1 outputs a close (OFF) to the alarm device 4, and the alarm device 4 stops the alert.

  Now, a determination is made regarding “FrameCount (Sunglasses, 2, 5)” which is an “individual condition” where “IO to be controlled” is ID2 and “control content” is open. Here, the determination unit 40 outputs “true” to “FrameCount (Sunglasses, 2, 5)” in the same manner as described above.

  Next, a determination is made regarding “FrameCount (Stay [1], 3, 5)”, which is an “individual condition” where “IO to be controlled” is ID2, and “control content” is open. Here, the determination unit 40 outputs “true” to “FrameCount (Stay [1], 3, 5)” in the same manner as described above.

Next, a determination is made regarding “TimeCount (Rove, 5000, 10000)” which is an “individual condition” where “IO to be controlled” is ID2 and “control content” is open. This individual condition means:
This is a condition for determining whether or not the total of the times when the characteristic keyword “Rove” outputs “true” in the latest target time of 10000 msec is the detection time of 5000 msec or more. From FIG. 11, it is in the area 1 in 1000 to 3000 and 4000 to 9000 msec in the latest 10,000 msec. Therefore, the determination unit 40 outputs “true” to “TimeCount (Rove, 5000, 10000)”.

  Finally, the “total condition” (judgment condition statement) here is “FrameCount (Sunglasses, 2,5)” and “FrameCount (Stay [1], 3,5)”, which are “individual conditions”, “TimeCount (Rove, 5000,10000)” is connected with the connection key of “and”. According to the determination result of each “individual condition”, it is “true” and “true” and “true”. Therefore, the determination unit 40 finally determines “FrameCount (Sunglasses, 2, 5) which is the“ total condition ”. “True” is output for “and FrameCount (Stay [1], 3,5) and TimeCount (Rove, 5000,10000)”. As a result, “IO to be controlled” outputs “ON” as “control content” to the output target of ID2.

  On the other hand, a determination is made regarding “FrameCount (Stay [2], 3, 5)” which is an “individual condition” where “IO to be controlled” is ID2 and “control content” is closed. This individual condition means whether or not the number of times the feature keyword “Stay [2]” is “true” (Stay [2] = 1) out of the last five frames is three or more. It is a condition to do. From FIG. 11, it is not in area 2 in the latest 5 frames. Accordingly, the determination unit 40 outputs “false” to “FrameCount (Stay [2], 3, 5)”. Since there is only one “individual condition” here, this corresponds to the “total condition”. The determination unit 40 finally outputs “false” to “FrameCount (Stay [2], 3, 5)” which is the “total condition”. As a result, “control IO” does not output “control content” closed (OFF) to the output target ID2.

  Based on the above, the determination condition sentence with “controlling IO” being ID2 and the determination result for the output target destination with “controlling IO” being ID2 will be considered. That is, the user uses this judgment condition sentence to notify the output target of ID2 (for example, the monitoring center 3) when the person wearing sunglasses is located in the area 1 and is still in a hurry. The determination file 32 is defined and created with a purpose. In the actual feature data, when a feature or action person satisfying this determination condition is detected, the security device 1 outputs open (ON) to the monitoring center 3, and the monitoring center 3 receives the notification. Is.

  After that, this person moves out of area 1 and moves to area 2 (for example, a building entrance), so that “FrameCount (Stay [1], 3,5)” is not satisfied, and “FrameCount (Stay [ 2], 3,5) ”, and conversely for“ FrameCount (Sunglasses, 2,5) and FrameCount (Stay [1], 3,5) and TimeCount (Rove, 5000,10000) ” Since “false” is output and “true” is output for “FrameCount (Stay [1], 3,5)”, the security device 1 outputs closed (OFF) to the monitoring center 3. . The monitoring center 3 can grasp that the person has moved out of the area 2 (for example, a building entrance), that is, outside the building.

  In summary, the user sounds the alarm device 4 when the person wearing sunglasses is in the area 1 according to the determination condition sentence. Further, when the person is suspiciously behaving suspiciously, in addition to the alarm device 4, the person is notified to the monitoring center 3. Also, when a person wearing sunglasses moves from the area 1, the alarm device 4 is released, and when the person moves further to the area 2, that is, when the person goes out of the building, The report is canceled.

  In the determination by the determination file 32, for the convenience of explanation, for example, the most recent specified number of frames such as “FrameCount (Sunglasses, 2, 10)”, “FrameCount (Stay [1], 3, 5)” is 10 It is explained using relatively small numbers. However, in the actual case, as described above, the camera 2a captures about 20 to 30 images (frames) per second and inputs them to the security device 1. Therefore, a slightly larger value can be used as the realistic most recent specified number of frames. More specifically, for example, “FrameCount (Sunglasses, 20, 100)”, “FrameCount (Stay [1], 30, 50)”, and the like can be used.

[System operation]
A system operation (information processing) in the security system 100 according to the present embodiment will be described. Although there are some points that have already been described above, in order to show the overall operation, the following will be repeated: (1) feature data generation step, (2) suspicious person determination step, (3) determination result The description is divided into three main output steps.

(1) Feature data generation step When the security device 1 starts security, an image and a detection signal are input from the camera 2a and the sensor 2b. The security device 1 performs an image analysis process every time an image or a detection signal is input, and generates feature data for each person ID (given for each person using a face authentication result or the like) from the result of the image analysis process. From time to time, it is stored and updated in the feature DB 31. This operation is repeated as long as the security device 1 keeps guarding.

  FIG. 12 is a flowchart illustrating the feature data generation process. This will be described below with reference to the drawings.

  S1: The input unit 10 of the security device 1 determines whether an image (and a detection signal) is input. When security is started, in the normal case, images and detection signals are successively input from the camera 2a and the sensor 2b.

  S2: The image analysis processing unit 20 performs image analysis processing on the input image (and detection signal). Here, first, it is detected whether or not a person is shown by image analysis and detection signals.

  S3: The image analysis processing unit 20 determines whether a person is detected. This is because when no person is detected, feature data does not have to be generated (landscape image of a frog).

  S4: When a person is detected, the image analysis processing unit 20 analyzes the input image (detection signal) by image analysis according to each feature item (FIG. 3), and the feature data. The feature data value for each item is detected. Then, one feature data is generated for one image (one frame) and one person.

  S5: Finally, the image analysis processing unit 20 totals the feature data for each person ID, for example, by arranging them in time series, and accumulates them in the feature DB 31. Specifically, a person is identified using a face authentication result or the like for the detected person. If the person has been detected in the past, the person ID issued to that person is used as the feature data (FIG. 3). [ID = xxxx]). If it is the first person that has not been detected in the past, a new person ID is issued to the person, and the person ID is added to the feature data ([ID = xxxx] in FIG. 3).

As described above, each time an image including a person is input, feature data (see, for example, FIG. 3) is generated for each person ID and stored in the feature DB 31 (see, for example, FIG. 2).
(2) Suspicious Person Determination Step When the security device 1 starts security, the determination unit 40 compares and collates the determination file 32 with feature data for each person, and the person is a person to be detected as a suspicious person. It is determined whether or not there is. As the timing of the determination process, for example, the determination can be performed at the timing when some person is detected based on the input image (and the detection signal), or the determination can be performed periodically at a predetermined interval.

  FIG. 13 is a flowchart for explaining a suspicious person determination process. This will be described below with reference to the drawings.

  S11: The determination unit 40 of the security device 1 determines whether or not there is a timing for performing the determination process. Here, the determination process is started at the timing when a person is detected based on the image (and the detection signal) input by the image analysis processing unit 20.

  S12: The determination unit 40 acquires the determination file 32 from the storage unit 30 and reads the contents. As described above, the determination file 32 is a file in which a feature to be detected as a suspicious person based on person characteristic data is defined as a determination condition. The determination file 32 is created in advance by the user and stored in the storage unit 30 prior to the determination process. Yes.

  S13: The determination unit 40 refers to the feature data as being aggregated for each person from the feature data DB. The target person is not the whole person in the feature data DB, but the person detected by the image analysis processing unit 20 based on the image (and the detection signal) input at that time. Then, as long as the same person continues to be detected (for example, as long as the same person stays in the building), the determination process is continued.

  S14: The determination unit 40 compares and collates the feature data of the detected person based on the determination file 32, and determines whether or not the person should be detected as a suspicious person. The specific determination method is as described above.

  S15: When the determination unit 40 determines that the person should be detected as a suspicious person, the output unit 50 instructs the monitoring center 3 and the alarm device 4 to output an alarm. That is, the output target corresponding to “IO to be controlled” is output as “control content” as open (ON) or closed (OFF).

  The determination unit 40 starts the determination process at the timing when a person is detected based on the image (and the detection signal) input by the image analysis processing unit 20, but immediately after the start, one image (frame) is detected at that moment. ) Only, and from the relationship with the above-described function (for example, FrameCount (x, y, z)), until the predetermined number of frames are imaged and sufficient feature data is accumulated, the determination file 32 Judgment processing based on cannot be performed. Therefore, when 20 to 30 images (frames) can be captured per second as the imaging capability of the camera 2a, the determination unit 40 stores 1 second from 1 second immediately after human detection, and accumulates feature data that can be subjected to the determination process. The determination process may be started after waiting.

(3) Determination result output step When the security device 1 detects a suspicious person based on the determination file 32, the security device 1 responds to the detection of the suspicious person. That is, when a suspicious person is detected, a report to the monitoring center 3 and an output to the alarm device 4 are performed.

  FIG. 14 is a flowchart for explaining determination result output processing.

  S21: The output unit 50 of the security device 1 determines whether or not an output instruction has been input from the determination unit 40. The output instruction here is an instruction to output “ON” or “OFF” as “control content” to the output target corresponding to “IO to be controlled” (S15).

  S22: The output unit 50 performs control according to the output instruction. For example, a report is output to the monitoring center 3 or this is canceled. Further, for example, an alarm is sounded or stopped for the alarm device 4.

[Specific setting method]
The security device 1 according to the present embodiment has been described above. Here, the setting method in the case of installing the security apparatus 1 in a certain installation place is mentioned. Specifically, a maintenance / management application (referred to as a maintenance application) is provided for the security device 1, and the field engineer operates the security device 1 directly or via the terminal PC of the field technician. The setting operation can be easily performed using the security device 1, and the operation of the security device 1 can be started. This will be described below.

  FIG. 15 shows a setting screen example (part 1) of the maintenance application. For example, when installing the security device 1 in a waiting room of a hospital, a field engineer can perform basic settings using a setting screen of a maintenance application.

  In the map section 1401 on the right side of the screen, a basic drawing of an installation destination, an installation position of an object such as a chair or a shelf, a camera position, and an imaging area (detection area) of the camera are set. In the basic information section 1402 at the lower left of the screen, various camera setting information, for example, parameters such as the camera name, position, orientation, depression angle, rotation angle, and area are input. A video image captured by the camera is displayed on the video display unit 1403 at the upper left of the screen. When the camera moves according to the parameters of the basic information unit 1402, the camera of the map unit 1401 also moves.

  Here, for example, the first camera of the two cameras is installed near the reception counter. A second camera will be installed near the entrance. The field engineer adjusts the position of the camera on the setting screen of the maintenance application, and when the position is determined, sets the name of the imaging range and designates the corresponding area range from the map unit 1401. Specifically, for example, each of “area name” is input as “area 1 (acceptance)”, and a drag operation is performed on the map unit 1401 to set the range of area 1 corresponding to the input “area name”. Designate (in the figure, the elliptical range of area 1). When the area range is determined, the range is determined by pressing the “range determination” button.

  FIG. 16 shows a setting screen example (part 2) of the maintenance application. Depending on the purpose of use, a field engineer, a hospital system administrator, or a security officer can set determination conditions (create a determination file 32) using the setting screen of the maintenance application. Here, this setting screen is used to set (create) determination conditions corresponding to the determination file 32 in FIG.

  Specifically, when a person wearing sunglasses is in area 1 (reception), alarm 4 is sounded. Thereafter, when the person moves out of the area 1, the alarm device 4 is released. In addition, if the person wearing sunglasses is located in area 1 and is still a little intimidated, the monitoring center 3 is notified. Thereafter, when the user leaves the hospital by moving to the area 2 (entrance), the reporting to the monitoring center 3 is stopped.

  In the figure, an “output destination” 1501 is an item for setting an output target destination to be controlled when the determination condition is satisfied. Here, it can be selected from, for example, 1: alarm (alarm), 2: report (monitoring center).

  The “state at the time of detection” 1502 indicates the content of control for the output target to be controlled when the determination condition is satisfied. “Open” means ON, and “Close” means OFF. In normal cases, it is set to “open”. For example, when a suspicious person is detected, an alarm is sounded (that is, “open”).

  A “detection target item” 1503 is an item for setting a feature (or action) of a person to be detected. For the selected branch, for example, feature expressions easy to understand for the user are prepared in advance, such as “wearing sunglasses”, “leaving away from place”, “suspicious behavior”, and the like. Of course, in the actual determination file 32, a determination condition sentence is generated using a feature keyword corresponding to each feature expression and a combination thereof. In the following, feature expression related to suspicious person detection and determination conditions corresponding to the feature expression will be exemplified using the function FrameCount. Note that these feature expressions that are easy to understand for the user are assigned a set of open and closed to one feature expression. Accordingly, if one feature expression is selected from the choices of “detection target item” 1503, one set of “judgment condition sentence” is automatically generated.

"Sunglasses"
Open FrameCount (Sunglasses, y, z) and FrameCount (Stay [x], y, z)
Closed Not (FrameCount (Stay [x], y, z))
"Suspicious person":
Open FrameCount (Sunglasses, y, z) and FrameCount (Stay [x], y, z) and FramCount (Rove, y, z)
Closed FrameCount (Stay [x], y, z)
“Location” 1504 is an item for setting the location and position of a person to be detected. Here, it is possible to select, for example, from area 1 (reception) or area 2 (entrance) from the imaging range of the two cameras.

  “Sensitivity” 1505 is an item for setting the sensitivity (accuracy) of detection. Here, for example, it can be selected from 1 (high) to 5 (low). In the actual determination file 32, in the case of the function FrameCount, this sensitivity is reflected in the nearest specified number of frames and the designated number of times when the feature keyword is “true”. For example, in the case of sensitivity “2”, the number of detections decreases as the sensitivity increases, and can be reflected as “FrameCount (Sunglasses, 2, 5)”. For example, when the sensitivity is “1”, it can be reflected as “FrameCount (Sunglasses, 1, 5)”.

  “Pop-up” 1506 is an item for setting ON / OFF of a function that informs the user of a pop-up when the determination condition is satisfied on the monitor screen of the security device 1, for example.

  “Invalid” 1507 is an item for setting validity / invalidity of the setting line. A line (judgment condition) for which the check box is checked is treated as not present.

  As described above, the determination file creation unit 60 of the security device 1 according to the present embodiment can easily create the determination file 32 via the maintenance application setting screen. In particular, it is difficult for a hospital system administrator or security officer who does not have specialized knowledge to set or change the judgment conditions. By using such a maintenance application setting screen, the judgment conditions can be easily set. Can be set (creation of the determination file 32).

  Needless to say, the determination file 32 can be created without using the setting screen of the maintenance application according to the embodiment. For example, when it is desired to set a person who cannot be set only by the feature expression of “detection target item” 1503 on the setting screen of the maintenance application and the feature thereof, the field engineer determines the determination file 32 stored in the storage unit 30 of the security device 1 It is possible to create a more flexible and advanced determination condition by directly creating or editing the above.

[Summary]
As described above, the security device 1 according to the present embodiment detects a person having a feature defined in the determination file based on the determination file and the feature data aggregated for each feature item. In this determination file, feature keywords corresponding to individual feature items of the feature data can be used in combination, so that determination conditions for detection suitable for the user's intention can be defined and created flexibly. In addition, even if there is no knowledge from the setting screen, the determination condition for detection can be easily defined and created simply by selecting an abstract feature expression from the selection. In other words, when detecting a suspicious person, by defining and creating flexible detection criteria according to the installation location and purpose of use, it can be used universally regardless of the location or purpose of use. It is possible to provide a security device.

  Note that the present invention is not limited to such specific embodiments, and various modifications and changes can be made within the scope of the gist of the present invention described in the claims. In the above-described embodiment, an example in which the present invention is applied to a building is shown. However, it goes without saying that the present invention can be similarly applied to homes and apartments.

DESCRIPTION OF SYMBOLS 1 Security apparatus 2 Data input device 2a Camera 2b Sensor 3 Monitoring center 4 Alarm device 10 Input part 20 Image analysis process part 30 Storage part 40 Judgment part 50 Output part 60 Judgment file creation part 100 Security system

Claims (5)

First storage means for storing feature data including a plurality of types of feature data values indicating the features of a person in association with each person;
Second storage means for storing a determination file including a determination condition defined from a combination of characteristic keywords corresponding to a characteristic of a person indicated by the characteristic data value;
Based on the match comparison between the feature data value of the feature data and the determination condition of the determination file, it is determined whether or not the determination condition is satisfied, and when the determination condition is satisfied, the person associated with the feature data Determining means for detecting a person as a detected person;
An output means for outputting a control signal for the output destination when the detected person is detected;
The determination file includes a determination condition defined by combining a plurality of different types of feature keywords using a connection key, an output destination that outputs person detection when it is determined that the determination condition is satisfied, Configured with the contents of the control signal for the output destination,
Security device characterized by. A screen for creating the determination file is provided, and when one feature representation is selected from the feature representations indicating the characteristics of the person to be detected on the screen, the screen is associated with the feature representation in advance. A determination file that creates a combination of a plurality of different types of feature keywords using the feature keyword or the connection key and creates a determination file that includes a determination condition defined from the created feature keyword or the combination of feature keywords Creating means,
The determination file created by the determination file creation means is stored in the second storage means;
The security device according to claim 1. The determination means outputs a value of “true” or “false” for each feature keyword based on a match comparison with the feature data value corresponding to the feature keyword, and then outputs “true” or “false” When the value is combined with the connection key, it is determined whether or not the determination condition is satisfied. If the value is satisfied, the value “true” is output. If the value is not satisfied, the value “false” is output. To do,
The security device according to claim 1 or 2. The determination means is a case where a predetermined function is included in the determination condition, and for each characteristic keyword, the determination means includes the specified number of characteristic data specified by the first parameter of the predetermined function. Based on the comparison with the feature data value corresponding to the feature keyword, when the value of “true” or “false” is output, it is specified by the second parameter of the predetermined function among the specified number of feature data. When the specified number of feature data matches the feature keyword, a value of “true” is output,
The security device according to claim 3. Input means for inputting at least an image from the camera;
The image analysis processing is performed on the image, and image analysis means for collecting the feature data of the person analyzed in the image for each person,
Feature data aggregated for each person by the image analysis means is stored in the first storage means;
The security device according to claim 1, wherein:

Images