JP6720909B2 - Action detection device, method and program, and monitored person monitoring device - Google Patents

Description

The present invention relates to an action detection device, an action detection method, and an action detection program that detect a predetermined action of a monitored person who is a monitoring target to be monitored. The present invention also relates to a monitored person monitoring device using this behavior detection device.

Japan (Japan) is an aging society, more specifically the ratio of the population aged 65 and over to the total population due to the improvement of living standards, the improvement of hygiene environment and the improvement of medical care, etc. accompanying the post-war high economic growth. It is a super-aging society with an aging rate of over 21%. In 2005, the total population was 127.65 million, but the number of elderly people aged 65 and over was 255.6 million, while in 2020, the total population was 124.11 million. It is predicted that the elderly population will be approximately 3,456,000. In such an aging society, those requiring nursing care or those requiring nursing care (such as those requiring nursing care) due to illnesses, injuries, or aging, etc. It is expected to increase more than the above. And Japan is also a declining birthrate society with a total fertility rate of 1.43 in 2013, for example. For this reason, elderly care is also occurring in which elderly families (spouse, children, brothers) care for the elderly who need nursing or care.

Nurses and care recipients enter a facility such as a hospital or a welfare facility for the elderly (short-term nursing home for the elderly, nursing home for the elderly and nursing home for the elderly, etc. under Japanese law) to receive nursing and care. In such a facility, there is a problem that a nurse or the like suffers an injury due to a fall from a bed, a fall while walking, or the like, getting out of the bed and wandering around. It is necessary to respond to such a situation as soon as possible, and if such a situation is left untreated, it may develop into a larger problem. And caregivers check the safety and the situation by regularly patrol.

However, the increasing number of nurses and the like cannot keep up with the increasing number of nurses and the like, and the nursing industry and the nursing care industry are chronically short of labor. Further, the number of nurses, caregivers and the like decreases in the semi-night shift and night shift hours compared to the day shift hours, and the work load per person increases. Therefore, it is required to reduce the work load. In addition, the situation of the old-age care is not an exception even in the above-mentioned facilities, and it is often seen that an elderly nurse or the like takes care of an elderly nurse or the like. In general, physical strength deteriorates as the person gets older, so even if they are healthy, the burden of nursing, etc. will be heavier than that of young nurses, and their movements and judgments will be slow.

In order to reduce such labor shortage and the burden on nurses and the like, there is a demand for technology that complements nursing work and nursing care work. For this reason, in recent years, a monitored person monitoring device for monitoring (monitoring) a monitored person, such as a nurse requiring care, to be monitored has been studied and developed.

As one of such techniques, for example, there are a human body detection device disclosed in Patent Document 1 and an image processing device disclosed in Patent Document 2. The human body detection device disclosed in Patent Document 1 includes an imaging camera that captures an image of the floor surface of a monitored person's room, and an imaging range of the imaging camera that is separated and identified into a plurality of areas, and the monitored person exists in each area. Based on the detection area determination unit that detects the movement and the movement, and the detection result regarding the presence and movement of the person being monitored by the detection area determination unit, the area to be monitored is monitored in a predetermined area or over a predetermined area and its adjacent area. When a person detects that the person is in a predetermined state, a determination processing unit that issues a warning to the observer side that an abnormality has occurred is provided. Then, the detection area determination unit, as a kind of the plurality of areas, a bed area in which the bed on the floor of the monitored person exists, and a floor surface of the monitored person's room except at least this bed area. And a floor area that is the predetermined area is set, and the determination processing unit, upon detecting that the monitored person has moved from the bed area to the floor area, notifies the monitored person that the monitored person has left the bed. Report abnormalities.

The image processing device disclosed in Patent Document 2 determines whether a person leaves the bed or enters the bed based on an image capturing camera that performs image capturing and image information captured by the image capturing camera from diagonally above the room to below the room. An image processing device comprising: an image processing unit, wherein the image processing unit includes a bed area, which is an area occupied by a bed installed on an indoor floor surface, and an area of the floor surface in the image information. A floor area other than the area in which the bed is placed, and also, it is possible to identify the movement of the human body from the floor area to the bed area while passing through a boundary side that represents the lower side of the outer shape of the bed, When it is determined that the human body is moving in from the bed area to the floor area after passing through the boundary side, it is determined that the human body is out of bed.

On the other hand, in terms of confirmation of safety, a single person living alone is the same as the nursing-needed person or the like, and is a monitored person who is a monitoring target.

JP 2001-307246 A JP, 2002-230533, A

In the human body detection device disclosed in Patent Document 1, as shown in FIG. 1 of Patent Document 1, an imaging camera is arranged at a substantially central position of a ceiling so that the entire floor surface of a room can be imaged. I have decided to leave the bed. The image processing apparatus disclosed in Patent Document 2 determines whether the user enters the floor or leaves the floor, on the assumption that an image pickup camera is provided so that the image can be taken from the diagonally upper side of the room toward the lower side of the room.

By the way, the size (size) and shape of the room are actually various, and it is not always possible to arrange the camera at a position approximately at the center of the ceiling or at a position where the room can be photographed from obliquely above the room to below the room. Absent. Therefore, if the above assumption is omitted, the human body detection device disclosed in Patent Document 1 and the image processing device disclosed in Patent Document 2 have high accuracy in the determination of the leaving of the bed and the determination of each of the entrance and the leaving. Will fall.

The present invention has been made in view of the above circumstances, and an object thereof is action detection capable of detecting a predetermined action of a person to be monitored with high accuracy regardless of the position where the camera is placed. An apparatus, an action detection method and an action detection program, and a monitored person monitoring device using the action detection device.

In the action detection device, the action detection method, the action detection program, and the monitored person monitoring device according to the present invention, an image of the monitored person being monitored is captured from above, and a predetermined action in the monitored person is performed. From among a plurality of different action detection algorithms for detecting the action, the action detection algorithm is selected based on the positional relationship between the camera that captured the image and a set region that is a predetermined region in the image, and this selection A predetermined action in the monitored person is detected on the basis of the image by the action detection algorithm.

INDUSTRIAL APPLICABILITY The behavior detection device, the behavior detection method, the behavior detection program, and the monitored person monitoring device according to the present invention can detect a predetermined behavior of the monitored person with higher accuracy, regardless of the installation position of the camera. ..

The above, and other objects, features and advantages of the present invention will be apparent from the following detailed description and the accompanying drawings.

It is a figure which shows the structure of the to-be-monitored person monitoring system in embodiment. It is a figure which shows the structure of the sensor apparatus in the to-be-monitored person monitoring system of embodiment. It is a figure for demonstrating the setting area|region and action detection line which are used by the said sensor apparatus. It is a figure for demonstrating the positional relationship of the camera with respect to the said setting area. It is a figure for demonstrating the method of calculating|requiring the positional relationship of the said camera used with the said sensor apparatus. It is a flowchart of the main routine which shows the action detection algorithm for Xn used by the said sensor apparatus. 7 is a flowchart of a subroutine showing an Xn wakeup determination algorithm in the Xn action detection algorithm. It is a flowchart of the subroutine which shows the leaving determination algorithm for Xn in the action detection algorithm for Xn. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method which are used by the action detection algorithm for direct above used when the positional relationship of the said camera is immediately above. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method used with the front action detection algorithm used when the positional relationship of the said camera is this side. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method used with the horizontal action detection algorithm used when the positional relationship of the said camera is horizontal. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method used with the back action detection algorithm used when the positional relationship of the said camera is the back. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method used by the action detection algorithm for diagonal fronts used when the positional relationship of the said camera is diagonally front. It is a figure for demonstrating the wake-up determination method and leaving-bed determination method used by the action detection algorithm for diagonal depths used when the positional relationship of the said camera is diagonal depths. 6 is a flowchart of a main routine showing the operation of the sensor device. 16 is a flowchart of a subroutine showing an action detecting operation used in the main routine shown in FIG. 15. It is a figure for demonstrating the calculation method of the said weight when the positional relationship of the said camera is represented by each weight of several basic positional relationship.

An embodiment of the present invention will be described below with reference to the drawings. In addition, in each of the drawings, the components denoted by the same reference numerals indicate the same components, and the description thereof will be appropriately omitted. In the present specification, reference numerals without suffixes are used for generic names, and reference numerals with suffixes are used when referring to individual configurations.

The person-to-be-monitored device includes an action detection unit that detects a preset predetermined action in a monitored subject, and a notification unit that externally notifies the predetermined action detected by the action detection unit. And with. Then, the behavior detection unit detects a predetermined behavior of the monitored person by a predetermined behavior detection algorithm based on an image of the monitored person taken from above the monitored person. Such a person-to-be-monitored device may be realized by being integrally configured as one device, or may be realized by a plurality of devices as a system. Then, when the monitored person monitoring device is realized by a plurality of devices, the behavior detection unit may be mounted on any of the plurality of devices. As an example, here, an embodiment of the monitored person monitoring apparatus will be described in the case where the monitored person monitoring apparatus is realized by a plurality of devices as a system. Even in the case where the monitored person monitoring device is integrally configured as one device, the monitored person monitoring device can be configured as in the following description. Further, as an example, a case will be described here in which the action detection unit is mounted on the sensor device SU described later together with the notification unit. However, another device in this system, for example, a management server device SV described later Also in the case of being mounted on the fixed terminal device SP or the mobile terminal device TA, the monitored person monitoring device can be configured in the same manner as the following description.

FIG. 1 is a diagram showing a configuration of a monitored person monitoring system according to the embodiment. FIG. 2 is a diagram showing a configuration of a sensor device in the monitored person monitoring system of the embodiment. FIG. 3 is a diagram for explaining a set area and an action detection line used in the sensor device. Note that FIG. 3 is also an example of an input screen for inputting these setting areas and action detection lines. FIG. 4 is a diagram for explaining the positional relationship of the camera with respect to the setting area. FIG. 5 is a diagram for explaining a method for obtaining the positional relationship of the cameras used in the sensor device. FIG. 5A shows a case where the image center is located in front of the action detection line in the setting area (bedding area), and FIG. 5B shows a case where the image center is located in the setting area (bedding area). FIG. 6 is a flowchart of a main routine showing an action detection algorithm for Xn used in the sensor device. Xn is the positional relationship of the cameras, and in the present embodiment, it is immediately above, in front, behind, lateral, diagonally forward, and diagonally deep. FIG. 7 is a flowchart of a subroutine showing an Xn wakeup determination algorithm in the Xn action detection algorithm. FIG. 8 is a flow chart of a subroutine showing a bed leaving determination algorithm for Xn in the action detection algorithm for Xn. FIG. 9 is a diagram for explaining the wakeup determination method and the leaving determination method used in the action detection algorithm for directly above, which is used when the positional relationship of the cameras is directly above. The left side of FIG. 9 shows an example of the image acquired by the camera, and the right side of FIG. 9 shows the region of the human body (human body region) extracted from the image shown on the left side of FIG. 9. FIG. 10 is a diagram for explaining a wake-up determination method and a wake-up determination method used in the front action detection algorithm used when the positional relationship of the cameras is in front. FIG. 11 is a diagram for explaining a wake-up determination method and a wake-up determination method used in a lateral action detection algorithm used when the positional relationship of the cameras is horizontal. FIG. 12 is a diagram for explaining a wake-up determination method and a wake-up determination method used in a back action detection algorithm used when the positional relationship of the cameras is deep. FIG. 13 is a diagram for explaining a wakeup determination method and a wakeup determination method used in a diagonally forward action detection algorithm used when the positional relationship of the cameras is diagonally forward. FIG. 14 is a diagram for explaining a wake-up determination method and a wake-up determination method used in a diagonally-behind action detection algorithm used when the positional relationship of the cameras is diagonally-backward. In FIGS. 10 to 14, the left side of the figure shows before waking up, the center of the figure shows after waking up, and the right side of the figure shows after waking up. 11 to 14, the upper part shows an example of the image acquired by the camera, and the lower part shows the human body region extracted from the image shown in the upper part.

The monitored person monitoring system MS, which is one example of the monitored person monitoring apparatus implemented as a system, has a monitored person (watched person) Ob (Ob-1) that is a monitored target (watched) to be monitored (watched). To Ob-4), a predetermined behavior set in advance is detected to monitor the monitored person Ob. For example, as shown in FIG. 1, one or a plurality of sensor devices SU (SU-1). ~ SU-4), a management server device SV, a fixed terminal device SP, and one or a plurality of mobile terminal devices TA (TA-1, TA-2), which are wired or wireless, LAN ( Local Area Network), a telephone network, a data communication network, and other networks (networks, communication lines) NW are communicably connected. The network NW may be equipped with repeaters such as repeaters, bridges, routers, and cross-connects that relay communication signals. In the example illustrated in FIG. 1, the plurality of sensor devices SU-1 to SU-4, the management server device SV, the fixed terminal device SP, and the plurality of mobile terminal devices TA-1 and TA-2 are wireless including an access point AP. A LAN (for example, a LAN conforming to the IEEE 802.11 standard) NW is communicably connected to each other.

The monitored person monitoring system MS is arranged at an appropriate place according to the monitored person Ob. The person to be monitored (person to be watched) Ob is, for example, a person who needs nursing due to illness or injury, a person who needs nursing care due to deterioration of physical ability, or a single person living alone. In particular, from the viewpoint of enabling early detection and early action, the monitored person Ob may be a person who needs to be detected when a predetermined inconvenient event such as an abnormal state occurs in the person. preferable. Therefore, the monitored person monitoring system MS is preferably installed in a building such as a hospital, a welfare facility for the aged, and a dwelling unit according to the type of the monitored person Ob. In the example illustrated in FIG. 1, the monitored-person monitoring system MS is installed in a building of a nursing care facility that includes a plurality of living rooms RM in which a plurality of monitored persons Ob live and a plurality of rooms such as a nurse station.

The sensor device SU is a device that has a communication function of communicating with other devices SV, SP, TA via the network NW, detects the monitored person Ob, and transmits the detection result to the management server device SV. This sensor device SU will be described in more detail later.

The management server device SV has a communication function of communicating with other devices SU, SP, TA via the network NW, and receives the detection result of the monitored person Ob and the image of the monitored person Ob from the sensor device SU. It is a device that manages information (monitoring information) about monitoring the monitored person Ob. When the management server device SV receives the detection result of the monitored person Ob and the image of the monitored person Ob from the sensor device SU, the management server device SV stores (records) the monitoring information regarding the monitoring of the monitored person Ob, and A communication signal (monitoring information communication signal) accommodating the monitoring information regarding the monitoring of the observer Ob is transmitted to the fixed terminal device SP and the mobile terminal device TA. Further, the management server device SV provides the client with data in response to a request from the client (the fixed terminal device SP, the mobile terminal device TA, etc. in the present embodiment). Such a management server device SV can be configured by, for example, a computer with a communication function.

The fixed terminal device SP has a communication function of communicating with other devices SU, SV, TA via the network NW, a display function of displaying predetermined information, an input function of inputting predetermined instructions and data, and the like. The user interface of the monitored person monitoring system MS by inputting a predetermined instruction or data to be given to the management server device SV or the mobile terminal device TA or displaying the detection result or image obtained by the sensor device SU. A device that functions as a (UI). Such a fixed terminal device SP can be configured by, for example, a computer with a communication function.

The mobile terminal device TA performs a communication function of communicating with other devices SV, SP, SU via the network NW, a display function of displaying predetermined information, an input function of inputting predetermined instructions and data, and a voice call. It is equipped with a call function to perform, and inputs predetermined instructions and data to be given to the management server device SV and the sensor device SU, and displays the detection results and images obtained by the sensor device SU by the notification from the management server device SV. It is a device that receives and displays the monitoring information regarding the monitoring of the person Ob to be monitored. Such a mobile terminal device TA can be configured by a portable communication terminal device such as a so-called tablet computer, smartphone, or mobile phone.

Next, the above-mentioned sensor device SU will be further described. As shown in FIG. 2, the sensor device SU includes, for example, a camera 1, a sound input/output unit 2, a control processing unit 3, a communication interface unit (communication IF unit) 4, and a storage unit 5.

The camera 1 is a device that is connected to the control processing unit 3 and generates an image (image data) under the control of the control processing unit 3. The camera 1 is located above the existing space so that the space to be monitored by the person Ob to be monitored (the existing space, the room RM of the installation location in the example shown in FIG. 1) can be monitored. It is arranged and images the location space as an imaging target from above, generates an image (image data) of the bird's-eye view of the imaging target, and outputs the image of the imaging target to the control processing unit 3. The camera 1 produces|generates a still image and a moving image in this embodiment. Since there is a high probability that such a camera 1 can image the entire monitored person Ob, it is planned that the head of the monitored person Ob is located in the bed BT such as a bed lying on the monitored person Ob. It is preferable to arrange the head so that the imaging target can be imaged right above the preset head position (usually, the pillow disposition position), but in reality, the size (size) or shape of the room It is arranged at an appropriate position, for example, above the ceiling or a wall, depending on the situation. The camera 1 is an example of an image acquisition unit that acquires an image of the monitored person Ob, which is a monitoring target, taken from above the monitored person Ob.

Such a camera 1 may be a device that generates an image of visible light, but in the present embodiment, it is a device that generates an image of infrared light so that the monitored person Ob can be monitored even in a relatively dark place. is there. Such a camera 1 is, for example, in the present embodiment, an imaging optical system that forms an infrared optical image of an imaging target on a predetermined imaging surface, and a light receiving surface is arranged so as to coincide with the imaging surface. And an image sensor that converts an infrared optical image of the imaging target into an electrical signal, and image data that is data representing an infrared image of the imaging target by performing image processing on the output of the image sensor. It is a digital infrared camera including an image processing unit for generating the image. In the present embodiment, the imaging optical system of the camera 1 is preferably a wide-angle optical system (a so-called wide-angle lens (including a fish-eye lens)) having an angle of view capable of imaging the entire living room RM in which the camera 1 is arranged. The camera 1 may be an infrared thermography device.

Note that the sensor device SU may further include a Doppler sensor in order to detect a microscopic movement abnormality as one of the predetermined actions of the monitored person. This Doppler sensor is a body motion sensor that transmits a transmission wave, receives a reflection wave of the transmission wave reflected by an object, and outputs a Doppler signal of a Doppler frequency component based on the transmission wave and the reflection wave. .. When the object is moving, since the frequency of the reflected wave is shifted in proportion to the moving speed of the object by the so-called Doppler effect, the difference between the frequency of the transmitted wave and the frequency of the reflected wave (Doppler frequency component) Occurs. The Doppler sensor 11 generates a signal of this Doppler frequency component as a Doppler signal and outputs it to the control processing unit 3. The transmission wave may be an ultrasonic wave, a microwave, or the like, but is a microwave in the present embodiment. Since the microwave can pass through the clothes and be reflected on the body surface of the monitored person Ob, the movement of the body surface can be detected even when the monitored person Ob is wearing clothes, which is preferable. Then, the control processing unit 3 obtains the body movement (up and down movement of the chest) of the chest associated with the breathing motion of the monitored person Ob based on the Doppler signal of the Doppler sensor by the action detection processing unit 36 described later, and the chest thereof. When the disturbance of the cycle of the body movement or the amplitude of the body movement of the chest that is equal to or less than a preset threshold value is detected, it is determined that the minute body movement is abnormal.

The sound input/output unit 2 is a circuit that is connected to the control processing unit 3 and that acquires an external sound and inputs it to the sensor device SU. It is a circuit for generating and outputting sound. The sound input/output unit 2 includes, for example, a microphone that converts acoustic vibration of sound into an electric signal, a speaker that converts electric signal of sound into the acoustic vibration of sound, and the like. The sound input/output unit 2 outputs an electric signal representing an external sound to the control processing unit 3, and also converts the electric signal input from the control processing unit 3 into acoustic vibration of sound and outputs the acoustic vibration.

The communication IF unit 4 is a communication circuit that is connected to the control processing unit 3 and performs communication under the control of the control processing unit 3. The communication IF unit 4 generates a communication signal containing the data to be transferred, which is input from the control processing unit 3, according to the communication protocol used in the network NW of the monitored person monitoring system MS, and generates the generated communication signal. It is transmitted to other devices SV, SP, TA via the network NW. The communication IF unit 4 receives communication signals from other devices SV, SP, TA via the network NW, extracts data from the received communication signals, and a format in which the extracted data can be processed by the control processing unit 3. Data and output to the control processing unit 3. The communication IF unit 4 further uses, for example, a Bluetooth (registered trademark) standard, an IrDA (Infrared Data Association) standard, a USB (Universal Serial Bus) standard, and the like to input and output data to and from an external device. An interface circuit for performing the above may be provided.

The storage unit 5 is a circuit that is connected to the control processing unit 3 and stores various predetermined programs and various predetermined data under the control of the control processing unit 3.

The various predetermined programs include, for example, a control processing program such as a monitoring processing program that executes information processing regarding monitoring of the monitored person Ob. The monitoring processing program includes an area setting processing program for performing processing for setting a predetermined area in the image as a setting area, and a plurality of predetermined predetermined values stored in advance for detecting a predetermined behavior of the monitored person Ob. An action detection line setting processing program that is used for at least one of the action detection algorithms and that performs a process for setting a predetermined line in the image as an action detection line, and obtains the positional relationship of the camera 1 with respect to the setting area. A positional relationship calculation program, an algorithm selection program that selects an action detection algorithm based on the positional relationship obtained by the positional relationship calculation program from among the plurality of action detection algorithms, and the action detection algorithm selected by the algorithm selection program in the camera 1. An action detection processing program that detects a predetermined action in the monitored person Ob based on the acquired image, a notification processing program that notifies the predetermined action detected by the action detection program to the outside, and a moving image captured by the camera 12, A voice call is made with the fixed terminal device SP or the portable terminal device TA by using the streaming processing program for delivering the moving image to the fixed terminal device SP or the portable terminal device TA by streaming, and the sound input/output unit 2. It includes a nurse call processing program and the like. The various kinds of predetermined data include, for example, data necessary for executing the above-mentioned programs such as the setting area and the behavior detection line, and data necessary for monitoring the monitored person Ob. included.

In order to store the setting area, the behavior detection line, and the plurality of behavior detection algorithms, the storage unit 5 functionally includes the setting area storage unit 51, the behavior detection line storage unit 52, and the behavior detection algorithm storage unit 53. Equipped with.

The set area storage unit 51 stores a predetermined area in the image as a set area. The set area is an area used in at least one of the plurality of behavior detection algorithms for detecting a predetermined behavior in the monitored person Ob, and in the present embodiment, the predetermined behavior is the monitored person. Since it is Ob wake-up and wake-up, in order to detect the presence or absence of the wake-up and the presence or absence of the wake-up, the setting area AR, for example, as shown as an area surrounded by four points PT1 to PT4, This is an area of bedding BT such as a bed in an image obtained by capturing the location space of the camera 1.

The action detection line storage unit 52 stores a predetermined line in the image as an action detection line. The action detection line is a line used for at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person Ob, and in the present embodiment, the predetermined action is the monitored target. Since the person Ob wakes up and leaves, in order to detect the presence or absence of the wake-up and the presence or absence of the wake-up, the action detection line AL is, for example, as shown as a line segment of two points PT2 and PT3 in FIG. It is one or more boundary lines in a setting area AR set as an area of the bedding BT that enters and leaves the bed BT.

The action detection algorithm storage unit 53 stores a plurality of different action detection algorithms in association with the positional relationship of the camera 1 that has captured the image with respect to the setting area AR.

In the present embodiment, the positional relationship of the camera 1 with respect to the setting area AR is classified based on the setting area AR, for example, as shown in FIG. 4, and the camera 1 is located directly above the setting area AR. , Front PO2 where the camera 1 is located above the setting area AR, back PO3 where the camera 1 is located above the setting area AR, lateral PO4, PO4 where the camera 1 is located above the setting area AR, and camera 1 Are diagonally forward PO5 and PO5 located diagonally above and above the setting area AR, and diagonally back PO6 and PO6 where the camera 1 is diagonally above and behind the setting area AR. The immediately above PO1, the front PO2, the back PO3, the side PO4, the diagonally front PO5, and the diagonally back PO6 also have a basic positional relationship described later.

In the present embodiment, a plurality of mutually different action detection algorithms are prepared in advance for each of the six immediately above PO1, the front PO2, the back PO3, the side PO4, the diagonal front PO5, and the diagonal back PO6, The plurality of action detection algorithms are the action detection algorithm for directly above used when the positional relationship of the camera 1 is directly above PO1, and the action detection algorithm for near side used when the positional relationship of the camera 1 is near PO2. Algorithm, back action detection algorithm used when the positional relationship of the camera 1 is the back PO3, horizontal action detection algorithm used when the positional relationship of the camera 1 is the side PO4, It includes an action detection algorithm for diagonal front used when the positional relationship is diagonal front PO5, and an action detection algorithm for diagonal depth used when the positional relationship of the camera 1 is diagonally backward.

The action detection algorithm for immediately above, for example, as shown in FIG. 9, has the first area S1 of the monitored person Ob outside the set area AR on the image, and the set area AR and the monitored person Ob on the image. This is an algorithm for detecting a predetermined action in the monitored person Ob based on the second area S2 where and are overlapped.

The front action detection algorithm, for example, as shown in FIG. 10, a predetermined action on the monitored person Ob based on the distance Wa between the action detection line AL on the image and the foot position FP of the monitored person Ob. Is an algorithm for detecting. This distance Wa corresponds to the length of the monitored person Ob that extends outside the action detection line AL.

For example, as shown in FIG. 11, the action detection algorithm for lateral use has a first boundary line BL1 and a set region located at a position closest to the camera 1 along the lateral direction of the set region AR on the image, as shown in FIG. Based on the third area S3 in which the monitored area Ob and the set area AR between the second boundary line BL2 located farthest from the camera 1 along the horizontal direction of AR overlap, the monitored person This is an algorithm for detecting a predetermined action in Ob.

The back action detection algorithm detects a predetermined action of the monitored person Ob based on the second area S2 in which the set area AR and the monitored person Ob overlap each other on the image, as shown in FIG. 12, for example. It is an algorithm to do.

For example, as shown in FIG. 13, the diagonal forward action detection algorithm includes a distance Wa between the action detection line AL and the foot position FP of the monitored person Ob on the image, and a set area on the image. A first boundary line BL1 located closest to the camera 1 along the horizontal direction of AR and a second boundary line BL2 located farthest from the camera 1 along the horizontal direction of the setting area AR. This is an algorithm for detecting a predetermined action in the monitored person Ob based on the third area S3 in which the set area AR and the monitored person Ob overlap with each other.

For example, as shown in FIG. 14, the action detection algorithm for oblique depths is set as the first boundary line BL1 located at a position closest to the camera 1 along the horizontal direction of the setting area AR on the image, as shown in FIG. Based on the third area S3 where the set area AR and the monitored person Ob overlap with the second boundary line BL2 located farthest from the camera 1 along the lateral direction of the area AR, the monitored area is displayed. It is an algorithm for detecting a predetermined action in the person Ob.

These six action detection algorithms for directly above, the action detection algorithm for the front, the action detection algorithm for the back, the action detection algorithm for the sideways, the action detection algorithm for the diagonally forward action, and the diagonally backward action detection algorithm The action detection algorithm is generally the same procedure, and assuming that Xn is directly above, in front, behind, sideways, diagonally forward and diagonally backward (Xn=directly above, near, far behind, sideways, diagonally forward, diagonally back), FIG. Through the follow chart shown in FIG.

More specifically, in the action detection algorithm for Xn, as shown in FIG. 6, first, the region of the human body (human body region) is extracted from the image (S1n), and then the wakeup determination process for Xn is executed. (S2n), and the bed leaving determination process for Xn is executed (S3n). The human body region is extracted by extracting the moving body region from the image acquired by the camera 1 by using a method such as a background difference method or an interframe difference method (moving body region=human body region). In the background subtraction method, a background image is obtained in advance and is stored in advance in the storage unit 5 as one of the various kinds of predetermined data, and the presence or absence of a moving object is detected from the difference image between the image generated by the camera 1 and the background image. If there is a moving body as a result of this determination, the area of this moving body is set as the moving body area. In the frame difference method, the presence or absence of a moving object is determined from the difference image between the image of the current frame and the image of the past frame (for example, the frame one frame before) generated by the camera 1, and as a result of this determination, the moving object is detected. If there is, the moving body area is set as the moving body area. On the contrary, the wakeup determination process S2n for Xn may be performed after the wakeup determination process S3n for Xn is performed.

In the wakeup determination process S2n for Xn, as illustrated in FIG. 7, first, it is determined whether the current state of the monitored person Ob regarding wakeup is before wakeup (S21n). The current state of getting up of the monitored person Ob is stored in the storage unit 5. For example, the current state of the monitored person Ob regarding getting up is stored in the storage unit 5 by a flag indicating a state regarding getting up (wakeup state flag). For example, when it is determined that the monitored person Ob is before waking up, this wake-up state flag is set to "0", and when it is determined that the monitored person Ob is after waking up, This wake-up state flag is set to "1". As a result of this determination, if it is not before wake-up, that is, if it is after wake-up (wake-up state flag=1, No), the wake-up determination process S2n for Xn is ended, and if it is before wake-up (wake-up state For flag=0, Yes), the following process S22n is executed. In step S22n, it is determined whether or not the wakeup determination condition for Xn is satisfied. This processing S22n will be further described later for each Xn. As a result of this determination, if the wakeup determination condition for Xn is satisfied (Yes), the notification processing unit 37 is notified of wakeup (S23n), and the current state of the monitored person Ob regarding wakeup is stored in the storage unit 5 as after wakeup. It is stored (wake-up state flag←1, S24n), and the wake-up determination process S2n for this Xn is ended. On the other hand, as a result of the determination, if the wakeup determination condition for Xn is not satisfied (No), the current state of the monitored person Ob regarding wakeup is stored in the storage unit 5 as being before wakeup (wakeup state flag ← 0, S25n ), the wakeup determination process S2n for Xn is completed.

In the bed leaving determination processing S3n for Xn, as shown in FIG. 8, first, it is determined whether or not the current state of the monitored person Ob regarding bed leaving is before bed leaving (S31n). The current state regarding leaving the monitored person Ob is stored in the storage unit 5. For example, the current state of the monitored person Ob regarding leaving the bed is stored in the storage unit 5 by a flag indicating the state regarding leaving the bed (leaving state flag). For example, when it is determined that the monitored person Ob is before leaving the bed, the bed leaving state flag is set to "0", and when it is determined that the monitored person Ob is after leaving the bed, This bed leaving state flag is set to "1". As a result of this determination, if it is not before leaving the bed, that is, if it is after leaving the bed (bed leaving state flag=1, No), this bed leaving determination processing S3n for Xn is ended, and if it is before leaving the bed (bed leaving state For flag=0, Yes), the following process S32n is executed. In the process S32n, it is determined whether or not the leaving determination condition for Xn is satisfied. This process S32n will be further described later for each Xn. As a result of this determination, when the leaving determination condition for Xn is satisfied (Yes), the leaving processing is notified to the notification processing unit 37 (S33n), and the current state of the monitored person Ob regarding leaving the bed is stored in the storage unit 5 as after leaving the bed. It is stored (bed leaving state flag ←1, S34n), and the bed leaving determination process S3n for Xn is ended. On the other hand, as a result of the determination, if the bed leaving determination condition for Xn is not satisfied (No), the current state regarding the bed leaving of the monitored person Ob is stored in the storage unit 5 as being before bed leaving (bed leaving state flag ← 0, S35n ), the bed leaving determination processing S3n for Xn is ended.

Next, the wakeup determination condition for Xn and the wakeup determination condition for Xn will be described for each Xn with reference to FIGS. 9 to 14. The wakeup determination condition for Xn and the wakeup determination condition for Xn are, for example, in the present embodiment, the first area S1 of the monitored person Ob outside the setting area AR on the image, the setting area AR on the image and the monitored person. Ob overlaps the second area S2, the first boundary line BL1 located closest to the camera 1 along the lateral direction of the setting area AR and the camera 1 along the lateral direction of the setting area AR. A third area S3 where the set area AR and the monitored person Ob overlap with the second boundary line BL2 located at the farthest position, and the action detection line AL on the image and the foot of the monitored person Ob. It is a condition based on one or more of the distance Wa from the position FP. Generally, the occurrence of getting up is detected as the first area S1 increases, and the occurrence of getting out of the bed is detected as the first area S1 further increases. Generally, the occurrence of getting up is detected as the second area S2 becomes smaller, and the occurrence of getting out of bed is detected as it becomes smaller. Generally, the occurrence of getting up is detected as the third area S3 becomes smaller, and the occurrence of getting out of bed is detected as it becomes smaller. In general, the rise of the floor is detected as the distance Wa increases, and the rise of the bed is detected as the distance Wa further increases.

More specifically, first, as shown in FIG. 9, the wakeup determination condition for immediately above used in the action detection algorithm for immediately above is that the first area S1 of the monitored person Ob outside the set area AR on the image is It is whether or not a preset first threshold value (upward determination threshold value for immediately above) Th1 has been exceeded. When the first area S1 exceeds the first threshold Th1 (S1>Th1), it is determined that the monitored person Ob has woken up, and the current state of the monitored person Ob regarding waking up is set to after waking up. On the other hand, when the first area S1 is less than or equal to the first threshold Th1 (S1≦Th1), it is determined that the monitored person Ob has not woken up, and the current state of the monitored person Ob regarding waking up is It is said that The first area S1 is obtained, for example, by counting the total number of pixels in the human body region outside the set region AR among the human body regions extracted in the process S1n. Then, as shown in FIG. 9, the second floor area S2 in which the set area AR on the image and the monitored person Ob overlap is set in advance as the bed leaving determination condition for directly above used in the action detection algorithm for immediately above. It is whether or not the second threshold value (threshold determination threshold for immediately above bed) Th2 is exceeded. If the second area S2 does not exceed the second threshold Th2, that is, if the second area S2 is less than the second threshold Th2 (S2<Th2), it is determined that the monitored person Ob has left the bed, and The current state regarding leaving the bed by the observer Ob is after leaving the bed. On the other hand, when the second area S2 is greater than or equal to the second threshold Th2 (S2≧Th2), it is determined that the monitored person Ob has not left the bed, and the current state regarding the leaving of the monitored person Ob is before leaving the bed. It is said that The second area S2 is obtained, for example, by counting the total number of pixels in a region where the set region AR and the human body region overlap, out of the human body region extracted in the process S1n. These first and second thresholds Th1 and Th2 are empirically preset, for example, by statistically processing a plurality of samples.

As shown in FIG. 10, the wake-up determination condition for the foreground used in the foreground action detection algorithm is preset with a distance Wa between the action detection line AL on the image and the foot position FP of the monitored person Ob. It is whether or not the third threshold value (front wakeup determination threshold value) Th3 has been exceeded. When the distance Wa is greater than or equal to the third threshold value Th3 (Wa≧Th3), it is determined that the monitored person Ob has woken up, and the current state of the monitored person Ob regarding waking up is set to after waking up (FIG. 10). See center). On the other hand, when the distance Wa does not exceed the third threshold Th3, that is, when the distance Wa is less than the third threshold Th3 (Wa<Th3), it is determined that the monitored person Ob is not awakened and The current state of getting up by the observer Ob is before going up (see the left side of FIG. 10). Regarding the distance Wa, for example, in the human body region extracted in the process S1n, the position closest to the floor surface is the foot position FP of the monitored person Ob, and the action detection line AL and the foot position of this monitored object Ob are set. It is obtained by counting the total number of pixels in one line along the vertical direction between the FP and the FP. Then, as shown in FIG. 10, whether the distance Wa for the foreground leaving determination condition used in the foreground action detection algorithm exceeds a preset fourth threshold value (foreward leaving bed determination threshold value) Th4. Is. When the distance Wa is greater than or equal to the fourth threshold Th4 (Wa≧Th4), it is determined that the monitored person Ob has left the bed, and the current state regarding the leaving of the monitored person Ob is after leaving the bed (FIG. 10). See the right side of). On the other hand, when the distance Wa does not exceed the fourth threshold value Th4, that is, when the distance Wa is less than the fourth threshold value Th4 (Wa<Th4), it is determined that the monitored person Ob has not left the bed, and The current state of leaving the observer Ob is assumed to be before leaving the bed (see the center of FIG. 10). The third and fourth threshold values Th3 and Th4 are empirically set in advance by statistically processing a plurality of samples, for example.

As shown in FIG. 11, the horizontal wake-up determination condition used in the horizontal action detection algorithm is that the first position is the closest to the camera 1 along the horizontal direction of the setting area AR on the image. A third area S3 in which the set area AR and the monitored person Ob overlap between the boundary line BL1 and the second boundary line BL2 located at the farthest position from the camera 1 along the lateral direction of the set area AR. Has exceeded a preset fifth threshold value (horizontal wakeup determination threshold value) Th5. When the third area S3 is less than or equal to the fifth threshold Th5 (S3≦Th5), it is determined that the monitored person Ob has woken up, and the current state of the monitored person Ob regarding waking up is set to after waking up ( (See the center of FIG. 11). On the other hand, when the third area S3 exceeds the fifth threshold Th5 (S3>Th5), it is determined that the monitored person Ob has not woken up, and the current state of the monitored person Ob regarding waking up is Before (see left side of FIG. 11). The third area S3 is, for example, the total number of pixels in a region where the set region AR sandwiched between the first boundary line BL1 and the second boundary line BL2 and the human body region overlap among the human body regions extracted in the process S1n. It can be obtained by counting. Then, as shown in FIG. 11, whether the third area S3 exceeds the preset sixth threshold (horizontal bed leaving determination threshold value) Th6 as the lateral bed leaving determination condition used in the horizontal action detection algorithm. It is not. When the third area S3 is equal to or smaller than the sixth threshold Th6 (S3≦Th6), it is determined that the monitored person Ob has left the bed, and the current state regarding the leaving of the monitored person Ob is after the bed leaving ( (See the right side of FIG. 11). On the other hand, when the third area S3 exceeds the sixth threshold value Th6 (S3>Th6), it is determined that the monitored person Ob has not left the bed, and the current state of the monitored person Ob regarding the leaving bed is Before (see center of FIG. 11). The fifth and sixth threshold values Th5 and Th6 are empirically preset by statistically processing a plurality of samples, for example.

As shown in FIG. 12, the wake-up determination condition for the back used in the behavior detection algorithm for the back is the seventh area in which the second area S2 in which the set area AR on the image and the monitored person Ob overlap is preset. It is whether or not the threshold (backward wakeup determination threshold) Th7 is exceeded. When the second area S2 is equal to or less than the seventh threshold value Th7 (S2≦Th7), it is determined that the monitored person Ob has woken up, and the current state of the monitored person Ob regarding waking up is set to “after waking up” ( (See the center of FIG. 12). On the other hand, when the second area S2 exceeds the seventh threshold value Th7 (S2>Th7), it is determined that the monitored person Ob has not woken up, and the current state of the monitored person Ob regarding waking up is Before (see left side of FIG. 12). Then, as shown in FIG. 12, whether the second area S2 exceeds the preset eighth threshold value (backward bed determination threshold value) Th8 as shown in FIG. It is not. When the second area S2 is equal to or less than the eighth threshold Th8 (S2≦Th8), it is determined that the monitored person Ob has left the bed, and the current state regarding the leaving of the monitored person Ob is after the bed leaving ( (See the right side of FIG. 12). On the other hand, when the second area S2 exceeds the eighth threshold Th8 (S2>Th8), it is determined that the monitored person Ob has not left the bed, and the current state of the monitored person Ob regarding the bed leaving is Before (see center of FIG. 12). The seventh and eighth thresholds Th7 and Th8 are empirically preset, for example, by statistically processing a plurality of samples.

As shown in FIG. 13, the wake-up determination condition for the diagonally forward action detection algorithm used in the diagonally forward action detection algorithm is preset with a distance Wa between the action detection line AL on the image and the foot position FP of the monitored person Ob. Whether or not the calculated ninth threshold value (first diagonal wake-up determination threshold value) Th9 has been exceeded, and on the image, it is located at a position closest to the camera 1 along the lateral direction of the setting area AR. A third area in which the set area AR and the monitored person Ob overlap between the first boundary line BL1 and the second boundary line BL2 located farthest from the camera 1 along the lateral direction of the set area AR. It is whether or not S3 exceeds a preset tenth threshold value (second diagonal front wakeup determination threshold value) Th10. When the distance Wa is greater than or equal to the ninth threshold Th9 and the third area S3 is less than or equal to the tenth threshold Th10 (W a ≧Th9 and S3 ≦Th10), it is determined that the monitored person Ob has woken up and is monitored. The current state of getting up by the person Ob is after getting up (see the center of FIG. 13 ). On the other hand, when the distance Wa does not exceed the ninth threshold Th9, that is, when the distance Wa is less than the ninth threshold Th9 and the third area S3 exceeds the tenth threshold Th10 (Wa<Th9 and S3>). At Th10), it is determined that the monitored person Ob has not woken up, and the current state of the monitored person Ob regarding waking up is pre-wakeup (see the left side of FIG. 13). As shown in FIG. 13, the diagonally-preceding wake-up determination condition used in the diagonally-preceding action detection algorithm is an eleventh threshold (first diagonal-previous wake-up determination threshold) Th11 in which the distance Wa is preset. And whether or not the third area S3 exceeds a preset twelfth threshold value (second oblique front wakeup determination threshold value) Th12. When the distance Wa is equal to or greater than the eleventh threshold Th11 and the third area S3 is equal to or less than the twelfth threshold Th12 (Wa≧Th11 and S3≦Th12), it is determined that the monitored person Ob has left the bed and is monitored. The current state regarding leaving the person Ob is assumed to be after leaving the bed (see the right side of FIG. 13 ). On the other hand, when the distance Wa does not exceed the eleventh threshold Th11, that is, when the distance Wa is less than the eleventh threshold Th11 and the third area S3 exceeds the twelfth threshold Th12 (Wa<Th11 and S3>). At Th12), it is determined that the monitored person Ob has not left the bed, and the current state regarding the leaving of the monitored person Ob is set to before leaving the bed (see the center of FIG. 13). The ninth to twelfth threshold values Th9 to Th12 are empirically preset, for example, by statistically processing a plurality of samples.

As shown in FIG. 14, the wake-up determination condition for the diagonal depth used in the behavior detecting algorithm for the diagonal depth is located at the position closest to the camera 1 along the lateral direction of the setting area AR on the image, as shown in FIG. The overlapping area of the set area AR and the monitored person Ob between the first boundary line BL1 and the second boundary line BL2 located at the farthest position from the camera 1 along the lateral direction of the setting area AR. It is whether or not the three areas S3 have exceeded a preset thirteenth threshold value (diagonal back rising determination threshold value) Th13. When the third area S3 is equal to or less than the thirteenth threshold Th13 (S3≦Th13), it is determined that the monitored person Ob has woken up, and the current state of the monitored person Ob regarding waking up is set to “after waking up” ( (See the center of FIG. 14). On the other hand, when the third area S3 exceeds the thirteenth threshold Th13 (S3>Th13), it is determined that the monitored person Ob has not woken up, and the current state of the monitored person Ob regarding waking up is Before (see left side of FIG. 14). Then, as shown in FIG. 14, the diagonally-backed-up bed leaving determination condition used in the diagonally-backward action detection algorithm is a 14th threshold (obliquely-backward leaving-up determination threshold) Th14 in which the third area S3 is preset. It is whether or not it exceeded. When the third area S3 is equal to or less than the fourteenth threshold Th14 (S3≦Th14), it is determined that the monitored person Ob has left the bed, and the current state regarding the leaving of the monitored person Ob is after leaving the bed ( (See the right side of FIG. 14). On the other hand, when the third area S3 exceeds the fourteenth threshold Th14 (S3<Th14), it is determined that the monitored person Ob has not left the bed, and the current state of the monitored person Ob regarding the leaving bed is Before (see center of FIG. 14). The thirteenth and fourteenth threshold values Th13 and Th14 are empirically set in advance by statistically processing a plurality of samples, for example.

The above-mentioned first to fourteenth threshold values Th1 to Th14 may be set in consideration of the size and action speed of the monitored person Ob on the image.

The above-described six action detection algorithms for directly above, the action detection algorithm for the front, the action detection algorithm for the back, the action detection algorithm for the lateral, and the diagonal action described above with reference to FIGS. 6 to 14. The previous action detection algorithm and the diagonal back action detection algorithm are stored in advance in the action detection algorithm storage unit 53.

The storage unit 5 includes, for example, a ROM (Read Only Memory) that is a non-volatile storage element, an EEPROM (Electrically Erasable Programmable Read Only Memory) that is a rewritable non-volatile storage element, and the like. The storage unit 5 includes a RAM (Random Access Memory), which serves as a working memory of the so-called control processing unit 3, which stores data and the like generated during execution of the predetermined program.

The control processing unit 3 controls each unit of the sensor device SU in accordance with the function of each unit, acquires an image of the monitored person Ob taken from above by the camera 1, and monitors the monitored object based on the acquired image. It is a circuit for detecting and notifying a predetermined action of the person Ob. The control processing unit 3 includes, for example, a CPU (Central Processing Unit) and its peripheral circuits. The control processing unit 3 executes the control processing program, whereby the control unit 31, the area setting processing unit 32, the action detection line setting processing unit 33, the positional relationship calculation unit 34, the algorithm selection unit 35, the action detection processing unit. 36, a notification processing unit 37, a streaming processing unit 38, and a nurse call processing unit 39 are functionally provided.

The control unit 31 controls each unit of the sensor device SU in accordance with the function of each unit, and controls the overall control of the sensor device SU.

The area setting processing unit 32 performs processing for setting a predetermined area in the image as the setting area AR. The action detection line setting processing unit 33 performs a process for setting a predetermined line in the image as the action detection line AL.

In the present embodiment, the control processing unit 3 receives at the communication IF unit 4 a communication signal (setting request communication signal) requesting the setting of each of the setting area AR and the action detection line AL from the fixed terminal device SP or the mobile terminal device TA. Then, the communication IF unit 4 sends back a communication signal (setting image communication signal) containing the image acquired by the camera 1. The fixed terminal device SP or the mobile terminal device TA displays, for example, a setting screen for inputting the setting area AR and the action detection line AL shown in FIG. The image contained in the area setting image communication signal is displayed on the setting screen. The user (monitored person, monitored person, etc.) uses, for example, a pointing device such as a mouse or tap operation on the touch panel to display the vertices PT1 to PT4 of the setting area AR and the action detection line AL on the setting screen. Both end points PT2 and PT3 are input. The input method may be input on each side of the setting area AR, input of the action detection line AL, or the like. Upon receiving the setting area AR and the action detection line AL, the fixed terminal device SP or the mobile terminal device TA sends the received communication signal (setting information communication signal) containing the setting region AR and the action detection line AL to the sensor device SU. Send. When the communication IF unit 4 receives the setting information communication signal, the area setting processing unit 32 extracts the setting area AR (4 vertices PT1 to PT4) contained in the received setting information communication signal and sets the setting in the storage unit 5. The setting area AR is set in the sensor device SU by storing it in the area storage unit 51, and the action detection line setting processing unit 33 stores the action detection line AL (both end points PT2, PT3 included in the received setting information communication signal. ) Is taken out and stored in the action detection line storage unit 52 of the storage unit 5 to set the action detection line AL in the sensor device SU. In the present embodiment, the communication IF unit 4 and the action detection line setting processing unit 33 correspond to an example of an action detection line reception unit that receives an action detection line from the outside and stores it in the storage unit.

The sensor device SU may further include an input unit 6 and a display unit 7, as shown by the broken line in FIG. The input unit 6 is a device that is connected to the control processing unit 3 and inputs, for example, various data necessary for monitoring the monitored person Ob to the sensor device SU, and includes, for example, a plurality of devices to which predetermined functions are assigned. An input switch, a mouse, etc. The display unit 7 is a device that is connected to the control processing unit 3 and displays data or the like input from the input unit 6 under the control of the control processing unit 3, and is a display device such as a CRT display, an LCD, an organic EL display, or the like. Is. A touch panel may be composed of the input unit 6 and the display unit 7. In the case of configuring this touch panel, the input unit 6 is a position input device that detects and inputs an operation position such as a resistance film type or an electrostatic capacitance type. In this touch panel, a position input device is provided on the display surface of the display unit 7, one or more input content candidates that can be input are displayed on the display unit 7, and the display position where the user displays the input content that the user wants to input is displayed. When is touched, the position is detected by the position input device, and the display content displayed at the detected position is input to the sensor device SU as the operation input content of the user. When such an input unit 6 and display unit 7 are provided, the control processing unit 3 displays on the display unit 7 a setting screen for inputting the setting area and the action detection line shown in FIG. 3, for example. The input unit 6 may receive the setting area AR and the action detection line AL. In this case, the image acquired by the camera 1 is displayed on the setting screen. The user uses, for example, a pointing device such as a mouse as the input unit 6 or performs a tap operation on the touch panel configured by the input unit 6 and the display unit 7, and the vertices PT1 to PT4 of the setting area AR on the setting screen. And both end points PT2 and PT3 of the action detection line AL are input. Upon receiving the setting area AR and the action detection line AL, the area setting processing unit 32 stores the setting area AR (4 vertices PT1 to PT4) received by the input unit 6 in the setting area storage unit 51 of the storage unit 5, The action detection line setting processing unit 33 stores the action detection line AL (end points PT2, PT3) received by the input unit 6 in the action detection line storage unit 52 of the storage unit 5. In such a case, the input unit 6 and the action detection line setting processing unit 33 correspond to another example of the action detection line reception unit that receives the action detection line from the outside and stores the action detection line in the storage unit.

The positional relationship calculation unit 34 calculates the positional relationship of the camera 1 with respect to the setting area AR. In the present embodiment, the positional relationship calculation unit 34 determines the positional relationship of the camera 1 with respect to the setting area AR in consideration of the action detection line AL. More specifically, for example, as shown in FIG. 5, the positional relationship calculation unit 34 determines the positional relationship of the camera 1 depending on the position of the image center CP with respect to the setting area AR and the action detection line AL. Then, the obtained positional relationship of the camera 1 is stored in the storage unit 5. As shown in FIG. 5, assuming that the left-right direction along the action detection line AL is the X direction and the front-rear direction orthogonal thereto is the Y direction, for example, as shown in FIG. 5A, the image center CP is less than the set area AR. In the Y direction (in front), there is no boundary of the set area AR between the action detection line AL and the image center CP, and the action detection line AL and the action detection line AL overlap at any position in the Y direction. In this case (that is, when the X coordinate value of the image center CP is included in the range from the X coordinate value of the one end PT2 to the X coordinate value of the other end PT3 in the action detection line AL), the positional relationship calculation unit 34 causes the camera As the positional relationship of 1, the front PO2 is calculated. Further, for example, as shown in FIG. 5B, when the image center CP is within the setting area AR, the positional relationship calculating unit 34 calculates the positional relationship of the camera 1 to be PO1 directly above. Further, for example, the image center CP is in the -Y direction (in front of) the setting area AR, there is a boundary of the setting area AR between the action detection line AL and the image center CP, and the action detection line AL. And the X-coordinate value of the image center is included in the range from the X-coordinate value of the one end PT2 to the X-coordinate value of the other end PT3 of the action detection line AL. ), the positional relationship calculation unit 34 calculates the back PO3 as the positional relationship of the camera 1. Further, for example, when the image center CP is in the X direction or −X direction (horizontal) from the setting area AR and overlaps with the setting area AR at any position in the X direction (that is, the image center CP is When the Y coordinate value is included in the range from the minimum Y coordinate value to the maximum Y coordinate value in the setting area AR), the positional relationship calculation unit 34 calculates lateral PO4 as the positional relationship of the camera 1. Further, for example, the image center CP is in the -Y direction (in front of) the setting area AR, there is no boundary of the setting area AR between the action detection line AL and the image center CP, and the action detection line AL. And the Y coordinate does not overlap at any position (that is, the X coordinate value of the image center CP is not included in the range from the X coordinate value of the one end PT2 to the X coordinate value of the other end PT3 on the action detection line AL). In this case, the positional-relationship calculation unit 34 calculates diagonally front PO5 as the positional relationship of the camera 1. Further, for example, the image center CP is in the -Y direction (in front of) the setting area AR, there is a boundary of the setting area AR between the action detection line AL and the image center CP, and the action detection line AL. And the Y coordinate does not overlap at any position (that is, the X coordinate value of the image center CP is not included in the range from the X coordinate value of the one end PT2 to the X coordinate value of the other end PT3 on the action detection line AL). In this case), the positional relationship calculation unit 34 calculates the diagonal relationship PO6 as the positional relationship of the camera 1. Note that, in FIG. 5A, since the action detection line AL is the line segment between the points PT2 and PT3 in the above description, the positional-relationship calculation unit 34 calculated that the positional relationship of the camera 1 is the front PO2. When the action detection line AL is a line segment between the points PT3 and PT4, the positional relationship calculation unit 34 calculates lateral PO4 as the positional relationship of the camera 1.

When the camera 1 has a tilting angle in a diagonally downward direction due to the camera 1 being disposed above the wall, the camera 1 is based on the tilting angle obtained by user input or automatic calculation based on an image by a known conventional means. The position directly below the camera 1 obtained by projecting the arrangement position of 1 on the floor surface (on the image) is obtained, and the obtained position directly below is the above-mentioned image center CP, and the positional relationship of the camera 1 is similar to the above. Is calculated.

The algorithm selection unit 35 selects an action detection algorithm from the plurality of action detection algorithms stored in the action detection algorithm storage unit 53 of the storage unit 5 based on the positional relationship obtained by the positional relationship calculation unit 34. The selection result is notified to the action detection processing unit 36. More specifically, in the present embodiment, the algorithm selection unit 35 determines the positional relationship of the camera 1 obtained by the positional relation calculation unit 34, and as a result of this determination, the camera determined by the positional relation calculation unit 34. In the case where the positional relationship of No. 1 is immediately above PO1, the above-mentioned action detection algorithm for immediately above is selected, and as a result of the determination, the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 is near PO2. If the above-mentioned front action detection algorithm is selected and the result of the determination is that the positional relationship of the camera 1 obtained by the positional relation calculation unit 34 is the back PO3, the back action detection described above is performed. When an algorithm is selected and the result of the determination is that the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 is horizontal PO4, the action detection algorithm for horizontal use described above is selected, and the result of the determination is When the positional relationship of the camera 1 obtained by the positional relationship calculating unit 34 is the front diagonal PO5, the above-described diagonal forward action detection algorithm is selected, and the result of the determination is obtained by the positional relationship calculating unit 34. When the positional relationship of the camera 1 is the diagonally rearward PO6, the behavioral detection algorithm for the diagonally rearward described above is selected, and the behavioral detection processing unit 36 is notified of the selection result.

The behavior detection processing unit 36 detects a predetermined behavior (presence or absence of waking up and presence or absence of wakeup in the present embodiment) of the monitored person Ob based on the image acquired by the camera 1 by the behavior detection algorithm selected by the algorithm selection unit 35. Then, the notification result is notified to the notification processing unit 37.

The notification processing unit 37 notifies the outside of the predetermined action (wake-up and wake-up in the present embodiment) detected by the action detection processing unit 36. More specifically, the notification processing unit 37 is information indicating the detected predetermined action (state, situation) (detection action information (in this embodiment, information indicating one or more of getting up and leaving)), Identifier information for identifying and identifying the monitored person Ob in which the predetermined action is detected (identifier information for identifying and identifying the sensor device SU detecting the monitored person Ob), and the predetermined The communication IF unit 4 generates a communication signal (monitoring information communication signal) containing an image or the like used for detecting the behavior of the above and transmits it to the management server device SV.

When there is a request for distribution of a moving image from the fixed terminal device SP or the mobile terminal device TA via the network NW and the communication IF unit 4, the streaming processing unit 38 makes a request for the fixed terminal device SP or the mobile terminal device. A moving image (for example, a live moving image) generated by the camera 1 is distributed to the TA by streaming reproduction via the communication IF unit 4 and the network NW.

When the nurse call processing unit 39 receives an input operation by a nurse call push button switch (not shown) that receives a nurse call from the monitored person Ob, the nurse call processing unit 39 receives the fixed terminal device SP or the fixed terminal device SP via the communication IF unit 4 and the network NW. This enables a voice call with the mobile terminal device TA.

In FIG. 1, as an example, four first to fourth sensor devices SU-1 to SU-4 are shown, and the first sensor device SU-1 is Mr. A who is one of the monitored persons Ob. The second sensor device SU-2 is arranged in the living room RM-1 (not shown) of Ob-1 and is arranged in the living room RM-2 (not shown) of Mr. B Ob-2 who is one of the monitored persons Ob. The third sensor device SU-3 is installed in the room RM-3 (not shown) of Mr. C. Ob-3 who is one of the monitored subjects Ob, and the fourth sensor device is the monitored subject. It is arranged in the room RM-4 (not shown) of Mr. D Ob-4 who is one of the Ob.

Next, the operation of this embodiment will be described. FIG. 15 is a flowchart of the main routine showing the operation of the sensor device. FIG. 16 is a flowchart of a subroutine showing the action detection operation used in the main routine shown in FIG.

In the monitored person monitoring system MS having such a configuration, each of the devices SU, SV, SP, and TA, when the power is turned on, executes necessary initialization of each unit and starts its operation. Further, in the sensor device SU, by executing the control processing program, the control processing unit 3 has the control unit 31, the area setting processing unit 32, the action detection line setting processing unit 33, the positional relationship calculation unit 34, and the algorithm selecting unit 35. The action detection processing unit 36, the notification processing unit 37, the streaming processing unit 38, and the nurse call processing unit 39 are functionally configured.

In FIG. 15, first, in the control processing unit 3, the area setting processing unit 32 sets the setting area AR (S41), and the action detection line setting processing unit 33 sets the action detection line AL (S42). More specifically, as described above, in the present embodiment, when the communication IF unit 4 receives the setting request communication signal from the fixed terminal device SP or the mobile terminal device TA, the control processing unit 3 causes the setting image communication to be performed. The signal is returned by the communication IF unit 4. The fixed terminal device SP or the mobile terminal device TA displays, for example, the setting screen shown in FIG. 3, and the user (monitoring person, monitored person, etc.) inputs the setting area AR and the action detection line AL to the setting screen. To do. Upon receiving the setting area AR and the action detection line AL, the fixed terminal device SP or the mobile terminal device TA transmits the setting information communication signal to the sensor device SU. When this setting information communication signal is received by the communication IF unit 4, the area setting processing unit 32 takes out the setting area AR contained in the received setting information communication signal and stores it in the setting area storage unit 51 of the storage unit 5. As a result, the setting area AR is set in the sensor device SU, and the action detection line setting processing unit 33 takes out the action detection line AL contained in the received setting information communication signal and takes the action detection line storage unit 52 of the storage unit 5. The action detection line AL is set in the sensor device SU by storing in the sensor device SU.

Next, the control processing unit 3 obtains the positional relation of the camera 1 with respect to the setting area AR by the positional relation calculating unit 34, and stores the obtained positional relation of the camera 1 in the storage unit 5 (S43). More specifically, as described above, in the present embodiment, the positional relationship calculation unit 34 determines the position of the camera 1 depending on the position of the image center CP with respect to the set area AR and the action detection line AL. Seek a relationship.

Next, the control processing unit 3 acquires an image of the monitored person Ob taken by the camera 1 from above the monitored person Ob. As a result, the image for one frame is acquired and input to the sensor device SU (S44).

Next, the control processing unit 3 executes, by the action detection processing unit 36, a later-described action detection process of detecting a predetermined action of the monitored person Ob based on the image acquired in the process S44 (S45), The process returns to step S44 in order to execute this action detection process on the image of the frame. That is, the control processing unit 3 repeatedly executes the processing S44 and the processing S45 to execute the action detection processing for each frame.

Next, the action detection process S45 will be further described. In this action detection process S45, in FIG. 16, the control processing unit 3 first causes the algorithm selection unit 35 to select a positional relationship among the plurality of action detection algorithms stored in the action detection algorithm storage unit 53 of the storage unit 5. An action detection algorithm is selected based on the positional relationship obtained by the calculation unit 34 (S51). More specifically, the algorithm selection unit 35 determines the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 by referring to the storage content of the storage unit 5.

As a result of this determination, when the positional relationship of the camera 1 obtained by the positional relationship calculating unit 34 in the process S43 is immediately above PO1 (immediately above), the algorithm selecting unit 35 selects the above-mentioned action detection algorithm for immediately above. Then, the action detection processing unit 36 is notified of this selection result. The action detection processing unit 36 that has received this notification executes processing S52-1. In this process S52-1, the action detection processing unit 36 causes the monitored person Ob to perform a predetermined action (in the present embodiment, based on the image acquired by the camera 1 by the action detection algorithm for immediately above selected by the algorithm selection unit 35). The presence/absence of waking up and the presence/absence of leaving the bed) are detected, the detection result is notified to the notification processing unit 37, and the action detection processing S45 is ended. That is, the behavior detection processing unit 36 determines whether or not the person Ob to be awakened and whether to leave the bed based on the image in each process (Xn=immediately above) according to the above-described flowcharts shown in FIGS. 6 to 8 for immediately above. When the presence/absence is determined and it is determined whether the user is getting up or leaving the floor, this is notified to the notification processing unit 37 as a detection result, and the action detection processing S45 is ended.

As a result of the determination in the process S51, when the positional relationship of the camera 1 obtained by the positional relationship calculating unit 34 in the process S43 is the front PO2 (front), the algorithm selecting unit 35 detects the front action detection described above. An algorithm is selected, and the action detection processing unit 36 is notified of the selection result. The action detection processing unit 36 that has received this notification executes processing S52-2. In this process S52-2, the action detection processing unit 36 selects the action detection algorithm for the front side selected by the algorithm selection unit 35 (that is, each process according to the above-described flowcharts shown in FIGS. 6 to 8 for the front side ( Xn=front)), a predetermined action (presence/absence of wake-up and presence/absence of wake-up) in the monitored person Ob is detected based on the image acquired by the camera 1, and the detection result is notified to the notification processing unit 37. Then, the action detection processing S45 is finished.

As a result of the determination in the process S51, when the positional relationship of the camera 1 obtained by the positional relationship calculating unit 34 in the process S43 is horizontal PO4 (horizontal), the algorithm selecting unit 35 causes the horizontal behavior detection described above. An algorithm is selected, and the action detection processing unit 36 is notified of the selection result. The action detection processing unit 36 that has received this notification executes processing S52-3. In this process S52-3, the action detection processing unit 36 determines the action detection algorithm for horizontal use selected by the algorithm selection unit 35 (that is, each process according to the above-described flowcharts shown in FIGS. 6 to 8 for horizontal use ( Xn=horizontal)), a predetermined action (presence/absence of wake-up and presence/absence of wake-up) of the monitored person Ob is detected based on the image acquired by the camera 1, and the detection result is notified to the notification processing unit 37. Then, the action detection processing S45 is finished.

As a result of the determination in the process S51, when the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 in the process S43 is the back PO3 (back), the algorithm selection unit 35 causes the back action detection described above. An algorithm is selected, and the action detection processing unit 36 is notified of the selection result. The action detection processing unit 36 that has received this notification executes processing S52-4. In this process S52-4, the action detection processing unit 36 selects the action detection algorithm for the depth selected by the algorithm selection unit 35 (that is, each process according to the above-described flowchart shown in FIGS. 6 to 8 for depth ( Xn=back)), a predetermined action (presence/absence of wakeup and presence/absence of wakeup) of the monitored person Ob is detected based on the image acquired by the camera 1, and the detection result is notified to the notification processing unit 37. Then, the action detection processing S45 is finished.

As a result of the determination in the process S51, when the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 in the process S43 is diagonally forward PO5 (obliquely forward), the algorithm selecting unit 35 uses the diagonally forward The action detection algorithm is selected and the action detection processing unit 36 is notified of the selection result. The action detection processing unit 36 that has received this notification executes processing S52-5. In this process S52-5, the action detection processing unit 36 follows the action detection algorithm for the diagonal front selected by the algorithm selection unit 35 (that is, according to the above-described flowcharts shown in FIGS. 6 to 8 for the diagonal front). A predetermined action (presence/absence of wake-up and presence/absence of wake-up in the present embodiment) of the monitored person Ob is detected based on the image acquired by the camera 1 in the processing (Xn=obliquely front)), and the detection result is notified by the processing unit. This is notified to 37 and this action detection processing S45 is ended.

As a result of the determination in the process S51, when the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 in the process S43 is the diagonal depth PO6 (diagonal depth), the algorithm selection unit 35 uses the above-described diagonal depth The action detection algorithm is selected, and the action detection processing unit 36 is notified of the selection result. The action detection processing unit 36 that has received this notification executes processing S52-6. In this process S52-6, the action detection processing unit 36 follows the action detection algorithm for diagonal depths selected by the algorithm selection unit 35 (that is, according to the above-described flowcharts shown in FIGS. 6 to 8 for diagonal depths). A predetermined action (presence/absence of wake-up and presence/absence of wake-up in the present embodiment) of the monitored person Ob is detected based on the image acquired by the camera 1 in the process (Xn=oblique depth)), and the detection result is notified by the processing unit. 37 is notified, and this action detection processing S45 is ended.

When receiving the notification of getting up by the processing S23n shown in FIG. 7 or the notification of leaving the bed by the processing S33n shown in FIG. A communication signal (monitoring information communication signal) containing the judgment result information (in the present embodiment, getting up and leaving) and monitoring information such as image data of a still image of the monitored person Ob is sent via the network NW to the management server device SV. Send to.

When the management server device SV receives the monitoring information communication signal from the sensor device SU via the network NW, the management server device SV stores the monitoring information such as the determination result information and the still image data contained in the monitoring information communication signal in its storage unit. Memorize (record) in. Then, the management server device SV transmits a monitoring information communication signal accommodating the determination result information and the monitoring information such as the image data of the still image to the terminal device (the fixed terminal device SP and the mobile terminal device TA in this embodiment). .. As a result, the state (situation) of the monitored person Ob is notified to the monitoring person such as a nurse or a caregiver via the terminal devices SP and TA.

When the fixed terminal device SP and the mobile terminal device TA receive the monitoring information communication signal from the management server device SV via the network NW, the fixed terminal device SP and the portable terminal device TA display the monitoring information contained in the monitoring information communication signal. With such an operation, the monitored person monitoring system MS detects each monitored person Ob by each sensor device SU, the management server device SV, the fixed terminal device SP, and the mobile terminal device TA, and detects each monitored person Ob. I'm watching.

As described above, the monitored person monitoring system MS which is an example of the monitored person monitoring apparatus, the action detection device, and the sensor device SU of the example using the action detection method and the action detection program are provided with a plurality of action detection algorithms. Since an action detection algorithm according to the positional relationship of the camera 1 with respect to the setting area AR is selected from the inside, and the predetermined action (wake-up and get-off in the present embodiment) of the monitored person Ob is detected by the selected action detection algorithm. The predetermined behavior of the monitored person Ob can be detected with higher accuracy regardless of the position where the camera 1 is arranged. Further, when the position of the bedding BT is changed after the camera 1 is provided and the action detection is started, the positional relationship of the camera 1 with respect to the bedding BT is changed. However, the monitored person monitoring system MS and the sensor device SU are described above. Can change the positional relationship of the camera 1 by changing the position of the bedding BT by changing the setting area AR stored in the storage unit 5 to the area of the changed bedding BT. Therefore, the monitored person monitoring system MS and the sensor device SU can detect the predetermined behavior of the monitored person Ob with higher accuracy regardless of the position of the bedding BT.

The monitored person monitoring system MS and the sensor device SU determine the positional relationship of the camera 1 as to the positional relationship between the camera 1 such as the immediately above PO1, the front PO2, the back PO3, the side PO4, the diagonally front PO5, and the diagonally back PO6. Since the camera 1 is provided with 34, it is possible to deal with the case where the camera 1 is arranged immediately above, in front, behind, sideways, diagonally in front, or diagonally in the back, and in these cases, the predetermined action of the monitored person Ob is more effective. It can be detected with high accuracy.

The monitored person monitoring system MS and the sensor device SU include the above-mentioned action detection algorithm for directly above, so that when the positional relationship of the camera 1 is immediately above PO1, the set area AR on the image is set. Whether the monitored person Ob wakes up by a relatively simple image process of obtaining the first area S1 of the outside monitored person Ob and the second area S2 in which the set area AR on the image and the monitored person Ob overlap each other. And it is possible to determine whether or not the person is out of bed.

Since the monitored person monitoring system MS and the sensor device SU include the front action detection algorithm described above with reference to FIG. 10, when the positional relationship of the camera 1 is the front PO2, the action detection line on the image is displayed. The presence/absence of waking up and the leaving of the monitored person Ob can be determined by relatively simple image processing of obtaining the distance Wa between the AL and the foot position FP of the monitored person Ob.

Since the monitored person monitoring system MS and the sensor device SU include the lateral action detection algorithm described above with reference to FIG. 11, when the positional relationship of the camera 1 is the lateral PO4, the set area is displayed on the image. A first boundary line BL1 located closest to the camera 1 along the lateral direction of AR and a second boundary line BL2 located farthest from the camera 1 along the lateral direction of the setting area AR. The presence/absence of rising and leaving of the monitored person Ob can be determined by relatively simple image processing of obtaining the third area S3 in which the set area AR and the monitored person Ob overlap each other.

Since the monitored person monitoring system MS and the sensor device SU include the back action detection algorithm described above with reference to FIG. 12, when the positional relationship of the camera 1 is the back PO3, the second area S2 is The presence/absence of waking up and the presence/absence of leaving the monitored person Ob can be determined by a relatively simple image processing of obtaining.

Since the monitored person monitoring system MS and the sensor device SU include the behavior detecting algorithm for diagonally forward movement described above with reference to FIG. 13, when the positional relationship of the camera 1 is diagonally forward PO5, the distance Wa and The presence or absence of waking up and the presence of leaving the monitored person Ob can be determined by relatively simple image processing of obtaining the third area S3.

Since the monitored person monitoring system MS and the sensor device SU include the behavior detecting algorithm for the diagonal depth described above with reference to FIG. 14, when the positional relationship of the camera 1 is the diagonal depth PO6, the third area is used. The presence/absence of waking up and the presence/absence of leaving of the monitored person Ob can be determined by relatively simple image processing of obtaining S3.

The monitored person monitoring system MS and the sensor device SU receive the action detection line AL from the outside and store it in the storage unit 5 as an example of the action detection line reception unit, and the communication IF unit 4 and the action detection line setting processing unit 33. Since it is provided, the action detection line AL can be easily input and set.

In the above embodiment, the positional relationship calculation unit 34 determines the positional relationship of the camera 1 depending on where the image center CP is with respect to the setting area AR and the action detection line AL. 3, the image center CP is also displayed on the display unit 7 in the image acquired by the camera 1 shown in FIG. The positional relationship between the area AR and the action detection line AL may be used.

Further, in the above-described embodiment, the algorithm selection unit 35 performs the action detection algorithm based on the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 from the plurality of action detection algorithms in the process S51 of the action detection process S45. However, the action detection algorithm may be selected at another timing after the positional relationship calculation unit 34 obtains the positional relationship of the camera 1. For example, the algorithm selection unit 35 may select the action detection algorithm immediately after the positional relationship calculation unit 34 obtains the positional relationship of the camera 1 in the above-described processing S43.

Further, in the above-described embodiment, the algorithm selection unit 35 selects the action detection algorithm based on the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34 from among a plurality of action detection algorithms for each predetermined period. In addition, when the request for the action detection algorithm selection process is received from the user, the algorithm selection unit 35 determines the position relation calculation unit 34 from the plurality of action detection algorithms at predetermined intervals by interrupt processing. An action detection algorithm based on the positional relationship of the camera 1 may be selected. In particular, it is preferable that the algorithm selection unit 35 selects the action detection algorithm by the interrupt process by the user after setting the setting area AR and setting the action detection line AL by the user. According to this, the rearrangement of the sensor device SU and the rearrangement of the bedding BT can be appropriately dealt with, and the predetermined behavior of the monitored person Ob can be detected with higher accuracy.

Further, in the above-described embodiment, the predetermined behavior of the monitored person Ob is, for example, getting up and leaving the bed, but is not limited to this. The predetermined behavior of the monitored person Ob may be another behavior as long as it is a behavior that can be detected based on an image captured from above the monitored person. For example, the predetermined behavior of the person Ob to be monitored may be protruding by rolling over or the like. The protrusion can be detected based on the head or arm extracted from the human body region by pattern matching using, for example, the pattern of the head or the pattern of the arm. For example, when the positional relationship of the camera 1 is directly above, the position of the head or arm during sleep can be detected by an action detection algorithm using a part area that exceeds the boundary line provided near the bed boundary. When the positional relationship is in front, it can be detected by the action detection algorithm using the part length that exceeds the boundary line corresponding to the bed height. The fall can be detected based on the head or trunk extracted from the human body region by pattern matching using a pattern of the head or a pattern of the trunk. For example, when the positional relationship of the camera 1 is directly above the PO1, it can be detected by an action detection algorithm for directly above using the difference in the size of the head or the size of the core due to getting up and leaving the bed, and for example, the positional relationship of the camera 1 However, in front PO2, it is possible to detect with a front action detection algorithm that uses the difference in the amount of vertical movement of the head or the amount of vertical movement of the trunk due to waking up or leaving the floor. In PO3, it is possible to detect with a back action detection algorithm that uses the difference in the area of the head and the area of the trunk visible on the action detection line AL associated with getting up and leaving the bed. In the lateral PO4, it is possible to detect by a lateral action detection algorithm that uses the difference in the left and right and up and down movement amounts of the head and the left and right and up and down movement amounts of the trunk due to getting up and leaving the bed. In the front PO5 diagonally, the above-mentioned action detection algorithms for front and side can be used for detection, and, for example, in the rear PO6 where the positional relationship of the camera 1 is diagonal, each of the above-mentioned back and side is detected. It can be detected using an action detection algorithm.

Further, in the above-described embodiment, the wakeup determination condition for Xn and the wakeup determination condition for Xn are one or more of the first area S1, the second area S2, the third area S3, and the distance Wa. However, it is not limited to this. The wakeup determination condition for Xn and the wakeup determination condition for Xn may be other conditions as long as they can be determined based on an image taken from above the monitored person. For example, the rising determination condition for Xn and the leaving determination condition for Xn may be a comparison of a ratio of the first area S1 and the second area S2 with a predetermined threshold Th. The threshold Th is empirically set in advance, for example, by statistically processing a plurality of samples in the same manner as described above. Further, for example, the wakeup determination condition for Xn and the wakeup determination condition for Xn may be the presence/absence of an operation across the action detection line AL. For example, when the positional relationship of the camera 1 is directly above PO1 or in front PO2, the wake-up is determined by detecting the motion of the human body region straddling the action detection line AL for the first time, and leaving the bed by detecting the motion of the entire human body region straddling the action detection line AL. When the positional relationship of the camera 1 is PO3, the wakeup is determined by detecting the motion of the human body region crossing the action detection line AL for the first time, and the wakeup is determined by detecting the motion of half of the human body region crossing the action detection line AL. Is determined.

Further, although the setting area AR is set by the user in the above-described embodiment, it may be automatically set. According to this, the setting area AR can be automatically set. For example, the area of the bedding BT is detected by extracting edges from the image acquired by the camera 1 using the edge filter, and the detected area of the bedding BT is automatically set as the setting area AR. Further, for example, the area of the bedding BT is detected from the image acquired by the camera 1 by pattern matching using the pattern of the bedding BT, and the detected area of the bedding BT is automatically set as the setting area AR. Further, for example, by using a sensor such as a carpet having a 3D sensor, a position sensor attached to a leg of a bed, and a pressure sensor laid under the bedding BT, an area of the bedding BT is detected, and the detected area of the bedding BT is detected. The area is automatically set as the setting area AR.

Further, in the above-described embodiment, the sensor device SU includes the communication IF unit 4 and the action detection line setting processing unit 33 as another example of the action detection line receiving unit, but the sensor device SU includes the control processing unit. 3 functionally further includes an action detection line calculation unit that obtains the action detection line AL based on the image acquired by the camera 1 and stores the obtained action detection line AL in the action detection line storage unit 52 of the storage unit 5. You may prepare. According to this, since the action detection line calculation unit is further provided, the action detection line AL can be automatically set. For example, the area of the bedding BT is detected by edge extraction or pattern matching from the image acquired by the camera 1, and the long side of the detected bedding BT is automatically set as the action detection line AL. Further, for example, similarly to the above, the area of the bedding BT is detected, the camera 1 as an example of the image acquisition unit acquires a plurality of images captured at different times, and the moving object area is a human body from the plurality of images. The frequency with which the detected human body region goes out of the region of the bedding BT is obtained for each side, and the side with the highest frequency in a predetermined period (setting period) is automatically set as the action detection line AL. To be done.

Further, in the above-described embodiment, the number of the action detection lines AL is one, but it may be plural. In this case, the positional relationship of the camera 1 is calculated for each behavior detection line AL, and the predetermined behavior in the person Ob to be monitored by the behavior detection algorithm based on the calculated positional relationship of the camera 1 for each behavior detection line AL. Is detected.

Further, in the above-described embodiment, the positional relationship calculating unit 34 determines the positional relationship of the camera 1 as the basic positional relationship of the six immediately above PO1, the front PO2, the back PO3, the side PO4, the diagonal front PO5, and the diagonal back PO6. One of them is calculated, and the algorithm selection unit 35 corresponds to the positional relationship of the camera 1 obtained by the positional relationship calculation unit 34, for directly above, foreground, behind, sideways, diagonally forward and diagonally backward. Although any one of the behavior detection algorithms is selected, the behavior detection algorithm may be gradually switched according to the position of the camera 1. In this case, for example, in the above-described embodiment, when the positional relationship calculation unit 34 classifies the positional relationship of the camera 1 that has captured the image with respect to the setting area AR into a plurality of different basic positional relationships, By obtaining the weight of each basic positional relationship based on the position of the camera 1 with respect to the setting area AR, the positional relationship of the camera 1 that has captured the image with respect to the setting area AR is obtained. Based on the plurality of basic positional relationships, a plurality of predetermined evaluation values for detecting a predetermined behavior of the monitored person Ob are calculated based on the plurality of basic positional relationships. The weighting action detection algorithm includes an algorithm for weighting and detecting the predetermined action in the monitored person Ob based on a result of comparison between the weighted result and a predetermined threshold, and the algorithm selecting unit 35 includes the plurality of action detections. The weighting action detection algorithm is selected from the algorithms based on the positional relationship obtained by the positional relationship calculating unit 34, and the action detection processing unit 36 weights each of the plurality of basic positional relationships obtained by the positional relationship calculating unit 34. Using the weighting action detection algorithm selected by the algorithm selection unit 35, the predetermined action in the monitored person Ob is detected based on the image. According to this, the positional relationship of the camera 1 with respect to the setting area AR is not represented by one basic positional relationship, but is expressed by weights of each of the plurality of basic positional relationships, and the plurality of basic positional relationships corresponding to the plurality of basic positional relationships are represented. The evaluation value is weighted by each weight of the plurality of basic positional relationships, and a predetermined behavior in the monitored person Ob is detected based on the weighted result. The predetermined behavior of the monitored person Ob can be detected with higher accuracy than in the case of detecting the predetermined behavior of the observer Ob.

FIG. 17 is a diagram for explaining a method of calculating the weight when the positional relationship of the cameras is represented by the weight of each of a plurality of basic positional relationships. 17A shows the setting area AR and the image center CP, and FIG. 17B shows the enlarged setting area ARE and the image center CP obtained by enlarging the setting area AR shown in FIG. 17A by two times. More specifically, the basic positional relationship is, for example, the immediately above PO1, the front PO2, the back PO3, the side PO4, the diagonally front PO5, and the diagonally back PO6, and the predetermined evaluation value is, for example, the above-mentioned first 1 area S1, 2nd area S2, 3rd area S3, and distance Wa. Then, as shown in FIG. 17, the positional-relationship calculating unit 34 doubles the setting area AR around the intersection of its diagonal lines (shown by broken lines in FIG. 17B) so that the setting area AR has a horizontal width W. The enlargement setting area ARE having the vertical width H is obtained. Next, the positional relationship calculation unit 34 obtains the distance p closest to the enlargement setting area ARE from the image center CP. In the example shown in FIG. 17, the image center CP is located on the lower right side in plan view in the setting area AR and the setting area AR is horizontally long. Therefore, the action detection line AL is determined to be the lower side of the setting area AR, and the image center The shortest distance p between the CP and the enlargement setting area ARE is the distance to the lower side of the enlargement setting area ARE. Then, the positional-relationship calculating unit 34 obtains a basic positional relation based on where the image center CP exists with respect to the setting area AR, the enlarged setting area ARE, and the action detection line AL, and the weighting of the obtained basic positional relationship. Ask for. In the example shown in FIG. 17, the image center CP is within the setting area AR, and both the action detection line AL and the side of the enlargement setting area ARE where the distance p from the image center is the shortest are below the image center CP. Therefore, the two basic positional relationships of the immediately above PO1 and the immediately preceding PO2 are calculated, the weight of the immediately preceding PO1 is p, and the weight of the immediately preceding PO2 is H/2−p. The above-mentioned wake-up determination condition for immediately above is a comparison between the first area S1 and the first threshold Th1, and the above-mentioned wake-up determination condition for this side is a comparison between the distance Wa and the third threshold Th3. The wakeup determination condition in the action detection algorithm is a comparison between V1=((Wa-Th3)*(H/2-p)+(S1-Th1)*p)*(H/2) and the fifteenth threshold Th15. .. For example, the fifteenth threshold Th15 is set to 0, and when the weighted result V1 is 0 or more (V1≧0), the behavior detection processing unit 36 determines that the monitored person Ob has woken up, and the weighted result When V1 is less than 0 (V1<0), it is determined that the monitored person Ob has not woken up. Further, the above-mentioned bed leaving determination condition for immediately above is a comparison between the second area S2 and the second threshold Th2, and the above-mentioned bed leaving determination condition for this side is a comparison between the distance Wa and the fourth threshold Th4. The bed leaving determination condition in the weighted behavior detection algorithm is V2=((Wa-Th4)*(H/2-p)+(S2-Th1)*p)*(H/2) and the 16th threshold Th16. Becomes For example, the sixteenth threshold Th16 is set to 0, and when the weighted result V2 is 0 or more (V2≧0), the behavior detection processing unit 36 determines that the monitored person Ob has left the bed, and the weighted result When V2 is less than 0 (V2<0), it is determined that the monitored person Ob has not left the bed.

Further, in this case, when the image center CP is located outside the enlargement setting area ARE, since the image capturing is performed by the camera 1 from a sufficient angle, the action detection algorithm for lateral use and the action detection algorithm for oblique back action described above. Etc. can be used as they are.

Further, in the above description, two basic positional relationships are selected and weighted, but three or more basic positional relationships may be selected and weighted.

Further, in the above-described embodiment, the sensor device SU includes the action detection unit and the notification unit, but the action detection unit and the notification unit may be included in the management server device SV, and the fixed terminal. It may be provided in the device SP or may be provided in the mobile terminal device TA. In such a case, a communication interface or the like that receives and acquires a communication signal containing an image captured from above the monitored person Ob and a Doppler signal from the sensor device SU via the network NW corresponds to an example of the image acquisition unit. Will be done.

Although the present specification discloses various aspects of the technique as described above, the main techniques thereof are summarized below.

The behavior detection device according to one aspect stores a monitored person as a monitoring target, an image acquisition unit that acquires an image captured from above the monitored person, and stores a predetermined area in the image as a setting area, A storage unit that stores a plurality of different action detection algorithms for detecting a predetermined action in the monitored person based on the image, a camera that captures the image from the plurality of action detection algorithms, and An algorithm selection unit that selects a behavior detection algorithm based on a positional relationship with a setting area, and a behavior detection process that detects a predetermined behavior in the monitored person based on the image with the behavior detection algorithm selected by the algorithm selection unit. And a section. Preferably, in the above-mentioned behavior detection device, the setting area is an area of bedding such as a bed (bedding area). Preferably, in the above-described action detection device, a positional relationship calculation unit that obtains the positional relationship between the camera that captured the image and the setting area depending on where the center of the image exists with respect to the setting area. Further prepare.

Such an action detection device selects an action detection algorithm according to the positional relationship between the camera and the setting area from among a plurality of action detection algorithms, and performs a predetermined action on the monitored person by the selected action detection algorithm. Since the detection is performed, it is possible to detect the predetermined behavior of the person to be monitored with higher accuracy regardless of the position where the camera is arranged. Further, when the position of the bedding is changed after the camera is provided and the action detection is started, the positional relationship of the camera with respect to the bedding changes, but the action detection device is configured to store the setting stored in the storage unit. By changing the region to the region of the bedding after the change, it is possible to deal with the change of the positional relationship of the camera due to the change of the position of the bedding. Therefore, the behavior detection device can detect a predetermined behavior of the monitored person with higher accuracy regardless of the position of the bedding.

In another aspect, in the above-described action detection device, the positional relationship is such that the camera is located immediately above the setting area, the camera is located above the setting area, and the camera is located above the setting area. The back is located above the back of the setting area, the side is above the setting area, the camera is located diagonally above the setting area, and the camera is located diagonally above the setting area. It is one of the diagonal depths located above the depths.

Such an action detection device can deal with the case where the camera is arranged directly above, in front, behind, sideways, diagonally forward, or diagonally deeply, and in these cases, it is possible to further improve the predetermined behavior of the monitored person. It can be detected with high accuracy.

In another aspect, in the above-described action detection device, the storage unit is used in at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, the image. A predetermined line in the inside is further stored as an action detection line, and the positional relationship calculation unit determines the positional relationship between the camera that has captured the image and the setting area in consideration of the action detection line.

According to this, it is possible to provide an action detection device that obtains the positional relationship of the cameras in consideration of the action detection line.

In another aspect, in these above-described action detection devices, the plurality of action detection algorithms include an action detection algorithm for directly above used when the positional relationship is directly above, and the action detection algorithm for immediately above is A predetermined area of the monitored person based on a first area of the monitored person outside the set area on the image and a second area of the set area on the image where the monitored person overlaps This is an algorithm that detects actions.

Such an action detection device is a relatively simple image processing that obtains the first area of the monitored person outside the set area on the image and the second area where the set area on the image and the monitored person overlap. It is possible to detect a predetermined action in the monitored person.

In another aspect, in the above-described action detection device, the storage unit is used in at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, the image. Further storing a predetermined line in the action detection line, the plurality of action detection algorithms include a foreground action detection algorithm used when the positional relationship is in front, the foreground action detection algorithm is Is an algorithm for detecting a predetermined action in the monitored person based on the distance between the action detection line on the image and the foot position of the monitored person.

Such an action detection device can detect a predetermined action in the monitored person by relatively simple image processing of obtaining the distance between the action detection line on the image and the toe position of the monitored person.

In another aspect, in these above-described action detection devices, the plurality of action detection algorithms include a horizontal action detection algorithm used when the positional relationship is horizontal, and the horizontal action detection algorithm is , On the image, located at a position closest to the camera along the lateral direction of the setting area and at a position farthest from the camera along the lateral direction of the setting area It is an algorithm that detects a predetermined action in the monitored person based on a third area where the set area and the monitored person overlap with the second boundary line.

Such an action detection device has the first boundary line located closest to the camera along the lateral direction of the setting area on the image and the farthest from the camera along the lateral direction of the setting area. The predetermined behavior of the monitored person can be detected by relatively simple image processing of obtaining a third area where the set area and the monitored person overlap with the second boundary line located at the position.

In another aspect, in these above-described action detection devices, the plurality of action detection algorithms include a back action detection algorithm used when the positional relationship is deep, and the back action detection algorithm is Is an algorithm for detecting a predetermined action in the monitored person based on a second area where the set area and the monitored person overlap on the image.

Such an action detection device can detect a predetermined action of the monitored person by relatively simple image processing of obtaining a second area where the set area on the image and the monitored person overlap.

In another aspect, in the above-described action detection device, the storage unit is used in at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, the image. Further storing a predetermined line as an action detection line, the plurality of action detection algorithms include a diagonal front action detection algorithm used when the positional relationship is diagonal front, the diagonal front action The detection algorithm is a distance between the action detection line and the toe position of the monitored person on the image, and, on the image, at a position closest to the camera along the lateral direction of the setting area. A third overlap between the set area and the monitored person between the first boundary line located and the second boundary line located farthest from the camera along the lateral direction of the setting area. It is an algorithm for detecting a predetermined action in the monitored person based on the area.

Such an action detection device is located at a position closest to the camera along the distance between the action detection line on the image and the toe position of the monitored person, and on the image in the lateral direction of the setting region. A third area in which the setting area and the monitored person overlap between a first boundary line and a second boundary line located at a position farthest from the camera along the lateral direction of the setting area, A predetermined action in the monitored person can be detected by relatively simple image processing of obtaining.

In another aspect, in the above-described action detection device, the storage unit is used in at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, the image. A predetermined line in the inside is further stored as an action detection line, and the plurality of action detection algorithms include an action detection algorithm for an oblique depth used when the positional relationship is an oblique depth, and the action for the oblique depth. The detection algorithm is such that, on the image, the first boundary line located closest to the camera along the lateral direction of the setting region and the furthest away from the camera along the lateral direction of the setting region. It is an algorithm for detecting a predetermined action in the monitored person based on a third area where the set area and the monitored person overlap with a second boundary line located at a position.

Such an action detection device has the first boundary line located closest to the camera along the lateral direction of the setting area on the image and the farthest from the camera along the lateral direction of the setting area. The predetermined behavior of the monitored person can be detected by relatively simple image processing of obtaining a third area where the set area and the monitored person overlap with the second boundary line located at the position.

In another aspect, in the above-described action detection device, in the case where the positional relationship calculation unit classifies the positional relationship between the camera that has captured the image and the setting area into a plurality of different basic positional relationships, By obtaining the weight of each of the plurality of basic positional relationship based on the position of the camera with respect to the setting area, to obtain the positional relationship between the camera that has captured the image and the setting area, the plurality of action detection algorithm, A plurality of predetermined evaluation values for detecting a predetermined action in the monitored person based on the image are obtained according to the plurality of basic positional relationships, and the plurality of basic positional relationships are obtained in the obtained plurality of evaluation values. Weighting is performed by weighting, and an algorithm for detecting a predetermined action in the monitored person based on a result of comparison between the weighted result and a predetermined threshold is included as a weighted action detection algorithm, and the algorithm selecting unit includes the plurality of actions. Select the weighting behavior detection algorithm based on the positional relationship obtained by the positional relationship calculation unit from the detection algorithm, the action detection processing unit, each of the plurality of basic positional relationship obtained by the positional relationship calculation unit The weighting action detection algorithm selected by the algorithm selection unit using the weight detects a predetermined action in the monitored person based on the image.

In such an action detection device, the positional relationship of the camera with respect to the set area is not represented by one basic positional relationship, but by the weight of each of the plurality of basic positional relationships, and a plurality of positions corresponding to the plurality of basic positional relationships are represented. The evaluation value is weighted by each of the plurality of basic positional relations, and a predetermined action in the monitored person is detected based on the weighted result, so that the action detection algorithm according to one basic positional relation Compared with the case of detecting a predetermined action of the monitored person, the predetermined action of the monitored person can be detected with higher accuracy.

In another aspect, the above-described action detection device further includes an action detection line reception unit that receives the action detection line from the outside and stores the action detection line in the storage unit.

Since such an action detection device further includes the action detection line reception unit, the action detection line can be easily input and set.

In another aspect, in the above-described action detection device, an action detection line calculation unit that obtains the action detection line based on the image acquired by the image acquisition unit and stores the obtained action detection line in the storage unit. Further prepare.

Since such an action detection device further includes the action detection line calculation unit, the action detection line can be automatically set.

An activity detection method according to another aspect is an image acquisition step of acquiring an image of a monitored person who is a monitoring target taken from above the monitored person, and for detecting a predetermined action in the monitored person. From a plurality of different behavior detection algorithms, an algorithm selection step of selecting a behavior detection algorithm based on the positional relationship between the camera that captured the image and a set area that is a predetermined area in the image, and the algorithm. An action detection process step of detecting a predetermined action in the monitored person based on the image by the action detection algorithm selected in the selection step.

An action detection program according to another aspect causes a computer to perform an image acquisition step of acquiring an image of a monitored person who is a monitoring target from above the monitored person, and a predetermined action in the monitored person. An algorithm selection step of selecting an action detection algorithm based on a positional relationship between a camera that has captured the image and a set region that is a predetermined region in the image, from among a plurality of different action detection algorithms for detecting. , A behavior detection processing step of detecting a predetermined behavior of the monitored person based on the image by the behavior detection algorithm selected in the algorithm selection step.

Such an action detection method and action detection program selects an action detection algorithm according to the positional relationship between the camera and the setting area from among the plurality of action detection algorithms, and the selected person in the monitored person uses the selected action detection algorithm. Since the predetermined action is detected, the predetermined action of the person to be monitored can be detected with higher accuracy regardless of the installation position of the camera. Further, when the position of the bedding such as the bed is changed after the camera is provided and the action detection is started, the positional relationship of the camera with respect to the bedding changes, but the action detection method and the action detection program are By changing the setting area stored in the storage unit, it is possible to deal with the change in the positional relationship of the camera due to the change in the position of the bedding. Therefore, the behavior detection method and the behavior detection program can detect a predetermined behavior of the monitored person with higher accuracy regardless of the position of the bedding.

A monitored person monitoring apparatus according to another aspect includes an action detection unit that detects a predetermined action in a monitored subject, and a notification unit that externally notifies the predetermined action detected by the action detection unit. And the action detecting unit includes any one of the above action detecting devices.

Since such a monitored person monitoring device includes any one of the above behavior detecting devices, it is possible to detect a predetermined behavior of the monitored person with higher accuracy regardless of the position where the camera is arranged. ..

This application is based on Japanese Patent Application No. 2015-118388 filed on June 11, 2015, the contents of which are included in the present application.

In order to represent the present invention, the present invention has been described above appropriately and sufficiently through the embodiments with reference to the drawings, but those skilled in the art can easily modify and/or improve the embodiments. It should be recognized that this is possible. Therefore, unless a modification or improvement carried out by a person skilled in the art is at a level that departs from the scope of rights of the claims recited in the claims, the modification or the improvement is covered by the scope of claims of the claim. Is understood to be included in.

According to the present invention, an action detection device, an action detection method and an action detection program, and a monitored person monitoring device can be provided.

Ob (Ob-1 to Ob-4) Monitored person NS (NS-1, NS-2) Monitored person (user)
MS Monitored person monitoring system SU Sensor device SV Management server device SP Fixed terminal device TA Mobile terminal device 1 Camera 3 Control processing unit 4 Communication interface unit (communication IF unit)
5 storage unit 31 control unit 32 area setting processing unit 33 action detection line setting processing unit 34 positional relationship calculation unit 35 algorithm selection unit 36 action detection processing unit 37 notification processing unit 51 set region storage unit 52 action detection line storage unit 53 action detection Algorithm storage

Claims (12)

An image acquisition unit that acquires an image of a monitored person who is a monitoring target from above the monitored person,
A storage unit that stores a predetermined area in the image as a setting area, and stores a plurality of different action detection algorithms for detecting a predetermined action in the monitored person based on the image,
From among the plurality of behavior detection algorithms, an algorithm selection unit that selects a behavior detection algorithm based on the positional relationship between the camera that captured the image and the setting area,
An action detection processing unit that detects a predetermined action in the monitored person based on the image with the action detection algorithm selected by the algorithm selection unit,
The plurality of behavior detection algorithms include a behavior detection algorithm for directly above used when the positional relationship is directly above,
The action detection algorithm for directly above is based on a first area of the monitored person outside the setting area on the image, and a second area where the setting area and the monitored person overlap on the image. Is an algorithm for detecting a predetermined action in the monitored person,
Behavior detection device. The positional relationship is directly above the camera located directly above the setting area, in front of the camera being located above the setting area, behind the camera being located above the setting area, and where the camera is It is one of a horizontal position above the setting region, a diagonal front position where the camera is diagonally above the setting region, and a diagonal depth position where the camera is diagonally above the setting region. ,
Behavior detection apparatus according to 請 Motomeko 1. Further comprising a positional relationship calculation unit that determines the positional relationship between the camera that captured the image and the setting area depending on where the center of the image exists with respect to the setting area ,
The storage unit further stores a predetermined line in the image used as at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, as an action detection line,
The positional relationship calculation unit obtains a positional relationship between the camera that has captured the image and the setting area in consideration of the action detection line,
The behavior detection device according to claim 2. The storage unit further stores a predetermined line in the image used as at least one of the plurality of action detection algorithms for detecting a predetermined action in the monitored person, as an action detection line,
The plurality of action detection algorithms include a front action detection algorithm used when the positional relationship is in front,
The foreground action detection algorithm is an algorithm for detecting a predetermined action in the monitored person based on the distance between the action detection line on the image and the toe position of the monitored person,
The behavior detection device according to claim 1. The plurality of action detection algorithms include a horizontal action detection algorithm used when the positional relationship is horizontal,
The action detection algorithm for horizontal use is, on the image, along the horizontal direction of the setting region, along a first boundary line located closest to the camera, and along the horizontal direction of the setting region. It is an algorithm for detecting a predetermined action in the monitored person based on a third area where the set area and the monitored person overlap with a second boundary line located farthest from the camera. The action detection device according to claim 1. The plurality of action detection algorithms include a back action detection algorithm used when the positional relationship is deep,
The behavior detection algorithm for depth use is an algorithm that detects a predetermined behavior of the monitored person based on a second area in which the set area and the monitored person overlap on the image.
The behavior detection device according to claim 1. In the case where the positional relationship calculation unit classifies the positional relationship between the camera that has captured the image and the setting area into a plurality of basic positional relationships that are different from each other, weights of the basic positional relationships are assigned to the setting area. By determining based on the position of the camera, the positional relationship between the camera that captured the image and the setting area,
The plurality of behavior detection algorithms, a plurality of predetermined evaluation values for detecting a predetermined behavior in the monitored person based on the image is obtained according to the plurality of basic positional relationship, the obtained plurality of evaluation values To each of the plurality of basic positional relationship weighted, including an algorithm for detecting a predetermined action in the monitored person as a weighted action detection algorithm based on a comparison result of the weighted result and a predetermined threshold value,
The algorithm selection unit selects the weighted behavior detection algorithm based on the positional relationship obtained by the positional relationship calculation unit from among the plurality of behavior detection algorithms,
The action detection processing unit uses a weighting action detection algorithm selected by the algorithm selection unit by using the weight of each of the plurality of basic positional relations calculated by the positional relation calculation unit to determine a predetermined value for the monitored person based on the image. To detect the behavior of
The behavior detection device according to claim 3 . Further comprising an action detection line receiving unit that receives the action detection line from the outside and stores the action detection line in the storage unit,
The behavior detection device according to claim 3. Further comprising an action detection line calculation unit that obtains the action detection line based on the image acquired by the image acquisition unit and stores the obtained action detection line in the storage unit.
The behavior detection device according to claim 3. An image acquisition step of acquiring an image of a monitored person who is a monitoring target from above the monitored person,
Action among a plurality of different action detection algorithms for detecting a predetermined action in the monitored person, based on the positional relationship between the camera that captured the image and a set region that is a predetermined region in the image An algorithm selection process of selecting a detection algorithm,
An action detection processing step of detecting a predetermined action in the monitored person based on the image by the action detection algorithm selected in the algorithm selection step,
The plurality of behavior detection algorithms include a behavior detection algorithm for directly above used when the positional relationship is directly above,
The action detection algorithm for directly above is based on a first area of the monitored person outside the setting area on the image, and a second area where the setting area and the monitored person overlap on the image. Is an algorithm for detecting a predetermined action in the monitored person,
Behavior detection method. On the computer,
An image acquisition step of acquiring an image of a monitored person who is a monitoring target from above the monitored person,
Action among a plurality of different action detection algorithms for detecting a predetermined action in the monitored person, based on the positional relationship between the camera that captured the image and a set region that is a predetermined region in the image An algorithm selection process of selecting a detection algorithm,
An action detection processing step of detecting a predetermined action in the monitored person based on the image by the action detection algorithm selected in the algorithm selection step,
An action detection program for executing
The plurality of behavior detection algorithms include a behavior detection algorithm for directly above used when the positional relationship is directly above,
The action detection algorithm for directly above is based on a first area of the monitored person outside the setting area on the image, and a second area where the setting area and the monitored person overlap on the image. Is an algorithm for detecting a predetermined action in the monitored person,
Behavior detection program. An action detection unit that detects a predetermined action in the monitored person who is the monitoring target,
A notification unit for notifying the outside of the predetermined action detected by the action detection unit,
The action detection unit includes the action detection device according to any one of claims 1 to 9 .
Monitored person monitoring device.

Images