JP7275863B2 - Image processing device, image processing device control method, and image processing device control program - Google Patents

Description

The present disclosure relates to an image processing device, a control method for an image processing device, and a control program for an image processing device.

2. Description of the Related Art Conventionally, there is known an image processing device that recognizes human behavior from an acquired image (see Patent Document 1, for example).

WO2016/181837

By the way, in this type of image processing apparatus, for example, from the viewpoint of early detection of an accident at the site of nursing care for the elderly, there is a demand for accurately detecting the occurrence of a falling state of a person.

Conventionally, in this type of image processing apparatus, the occurrence of a falling state of a person is detected from temporal changes in the posture of the person detected from time-series images. However, there are various situations in which this type of image processing apparatus is applied, and various peripheral objects exist around a person to be recognized.

Therefore, in the image processing apparatus according to the related art, there is a possibility that the person cannot be distinguished from the surrounding objects, and that the occurrence of the falling state of the person concerned is erroneously detected. In particular, in the image processing apparatus according to the related art, when a person leaves the imaging area of the imaging device through the door, or when the person is hidden behind an obstacle, the person and the door are integrated. may be recognized by As a result, the conventional image processing apparatus may recognize that the person has stopped at one point in the image, and may recognize that the person has fallen.

The present disclosure has been made in view of the above problems, and an image processing device, a control method for the image processing device, and an image processing device that can suppress erroneous detection related to the occurrence of a falling state of a person with a simple method. The purpose is to provide a control program.

The main disclosure that solves the above-mentioned problems is
an image acquisition unit that acquires time-series images generated by an imaging device that captures an image of a monitoring target area;
a falling state detection unit that detects the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected by the overturned state detection unit, a first image at the time when the occurrence of the overturned state is detected among the time-series images, and A second image at a time point a predetermined time before the detection start timing at which the occurrence of the overturned state is detected is extracted, and the overturned state detection unit detects the overturned state based on the amount of change in the first image with respect to the second image. an erroneous detection suppression unit that determines whether the result is an erroneous detection;
An image processing device comprising:
Also, in other aspects,
an image acquisition unit that acquires time-series images generated by an imaging device that captures an image of a monitoring target area;
a falling state detection unit that detects the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected by the overturned state detection unit, the occurrence of the overturned state is detected by the overturned state detection unit in the time-series images, and A first image after a predetermined continuation determination time from the detection start timing at which the occurrence of the overturning state is detected and a second image at a time point before the occurrence of the overturning state is detected are extracted, and the second image is extracted. an erroneous detection suppression unit that determines whether or not the detection result of the overturned state detection unit is an erroneous detection, based on the amount of change in the first image with respect to the image;
An image processing device comprising:

Also, in other aspects,
Acquiring time-series images generated by an imaging device that captures an image of a monitored area,
detecting the occurrence of a falling state of a subject existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the first image at the time when the occurrence of the overturned state is detected among the time-series images, and from the detection start timing at which the occurrence of the overturned state is detected. , extracting a second image at a point in time a predetermined time ago , and determining whether or not the occurrence of the falling state is an erroneous detection based on the amount of change in the first image with respect to the second image;
A control method for an image processing apparatus.
Also, in other aspects,
Acquiring time-series images generated by an imaging device that captures an image of a monitored area,
detecting the occurrence of a falling state of a subject existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the period during which the occurrence of the overturned state is detected in the time-series images and from the detection start timing at which the occurrence of the overturned state is detected extracting a first image after a predetermined continuation determination time and a second image at a point in time before the occurrence of the overturned state is detected, and based on a change of the first image with respect to the second image; , determining whether the occurrence of the falling state is an erroneous detection;
A control method for an image processing apparatus.

Also, in other aspects,
to the computer,
A process of acquiring time-series images generated by an imaging device that captures an image of a monitoring target area;
a process of detecting the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the first image at the time when the occurrence of the overturned state is detected among the time-series images, and from the detection start timing at which the occurrence of the overturned state is detected. , a process of extracting a second image at a point in time a predetermined time ago , and determining whether or not the occurrence of the overturned state is an erroneous detection based on the amount of change of the first image with respect to the second image;
is a control program for an image processing apparatus that executes
Also, in other aspects,
to the computer,
A process of acquiring time-series images generated by an imaging device that captures an image of a monitoring target area;
a process of detecting the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the period during which the occurrence of the overturned state is detected in the time-series images and from the detection start timing at which the occurrence of the overturned state is detected extracting a first image after a predetermined continuation determination time and a second image at a point in time before the occurrence of the overturned state is detected, and based on a change of the first image with respect to the second image; , a process of determining whether the occurrence of the falling state is an erroneous detection;
is a control program for an image processing apparatus that executes

According to the image processing device according to the present disclosure, it is possible to suppress erroneous detection related to the occurrence of a falling state of a person using a simple method.

A diagram showing an example of an action recognition system according to an embodiment. 1 is a diagram showing an example of the hardware configuration of an image processing apparatus according to an embodiment; FIG. 1 is a diagram showing an example of functional blocks of an image processing apparatus according to an embodiment; FIG. FIG. 4 is a diagram showing an example of a human region in an image detected by a human region detection unit according to one embodiment; FIG. 4 is a diagram showing an example of a monitoring target area to be determined for erroneous detection suppression by an erroneous detection suppression unit according to an embodiment; FIG. 4 is a diagram illustrating an example of determination processing of an erroneous detection suppressing unit according to an embodiment; FIG. 4 is a diagram illustrating an example of determination processing of an erroneous detection suppressing unit according to an embodiment; FIG. 5 is a diagram illustrating an example of a method of obtaining a change amount of an after image with respect to a before image by an erroneous detection suppressing unit according to one embodiment; 4 is a flow chart showing an example of the operation of an erroneous detection suppression unit according to an embodiment;

Preferred embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings. In the present specification and drawings, constituent elements having substantially the same functional configuration are denoted by the same reference numerals, thereby omitting redundant description.

[Overall Configuration of Action Recognition System]
Hereinafter, the overall configuration of an action recognition system U according to one embodiment will be described with reference to FIGS. 1 to 3. FIG.

FIG. 1 is a diagram showing an example of an action recognition system U according to this embodiment.

An action recognition system U according to this embodiment includes an image processing device 100 , an imaging device 200 and a communication network 300 .

The imaging device 200 is, for example, a general camera or a wide-angle camera, and AD-converts an image signal generated by an imaging element of the camera to generate image data. The image capturing apparatus 200 according to the present embodiment is configured to continuously generate image data in frame units and capture time-series image data (that is, moving image data).

The imaging device 200 is installed, for example, at an appropriate position in the room. Then, the imaging device 200 has a door B2 for the person B1 (corresponding to the “subject” of the present invention) to exit from the imaging area (here, the room) to the outside of the imaging area (here, the toilet adjacent to the room). is installed so that is reflected in the image.

The imaging device 200 transmits moving image data to the image processing device 100 via the communication network 300 .

The image processing device 100 is a device that determines the behavior of the person B1 appearing in the image based on the data of the image generated by the imaging device 200 and outputs the determination result.

FIG. 2 is a diagram showing an example of the hardware configuration of the image processing apparatus 100 according to this embodiment.

The image processing apparatus 100 includes, as main components, a CPU (Central Processing Unit) 101, a ROM (Read Only Memory) 102, a RAM (Random Access Memory) 103, an external storage device (for example, a flash memory) 104, a communication interface 105, and the like. is a computer with

For each function of the image processing apparatus 100, which will be described later, for example, the CPU 101 refers to control programs (eg, image processing programs) and various data (eg, image data) stored in the ROM 102, the RAM 103, the external storage device 104, and the like. realized by The RAM 102 functions, for example, as a data work area and a temporary save area.

However, each function of the image processing apparatus 100 is configured by a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field-Programmable Gate Array), or the like, instead of or together with the processing by the CPU 101. may be implemented by a digital arithmetic circuit.

FIG. 3 is a diagram showing an example of functional blocks of the image processing apparatus 100 according to this embodiment.

The image processing apparatus 100 includes an image acquisition section 10 , a human region detection section 20 , a falling state detection section 30 , and an erroneous detection suppression section 40 .

[Image Acquisition Unit]
The image acquisition unit 10 acquires image data (hereinafter abbreviated as “image”) generated by the imaging device 200 from the imaging device 200 . Then, the image acquisition unit 10 sequentially acquires time-series images that are continuously generated by the imaging device 200 .

The image acquired by the image acquiring unit 10 from the imaging device 200 is temporarily stored in the RAM 103 or the like for a predetermined period of time (for example, one minute).

[Human area detector]
The human region detection unit 20 performs predetermined arithmetic processing on the time-series images acquired by the image acquisition unit 10, and detects a region including the person B1 in the image (hereinafter referred to as a “human region”). .

FIG. 4 is a diagram showing an example of a human area in an image detected by the human area detection unit 20. As shown in FIG. In FIG. 4, Rall represents the entire image area of the image, and R1 represents the human area in the image.

For example, for each time-series image, the human region detection unit 20 detects a difference image from the image at the previous timing, and detects the human region R1 from the difference image. Then, the human region detection unit 20 sets, for example, a rectangular region including the region in which the person B1 is detected in the image as the human region R1.

However, other methods for detecting the region R1 by the human region detection unit 20 include methods such as trained neural networks, template matching, HOG (Histograms of Oriented Gradients) feature quantity, and SVM (Support Vector Machine). may

[Tumble state detector]
Based on the time-series images acquired by the image acquiring unit 10, the overturning state detection unit 30 detects the occurrence of the overturning state of the person B1 appearing in the images.

The fall state detection unit 30 determines whether or not the person B1 has fallen, based on the temporal change of the human region R1 detected by the human region detection unit 20, for example. At this time, the overturning state detection unit 30 determines whether or not the overturning state of the person B1 has occurred, for example, using the amount of change in the person region R1 per unit time. Specifically, the falling state detection unit 30 calculates the difference between the human regions R1 detected by the human region detection unit 20 every predetermined number of frames. It is determined that the overturned state of the person B1 has not occurred, and if the amount of change in the person region R1 per unit time is small, it is determined that the overturned state of the person B1 has occurred. This determination criterion is that when the amount of change in the person region R1 per unit time is large, the person B1 is not falling over because the person B1 is performing an action such as walking. When the amount of change is small, the person B1 does not move, and therefore the person B1 is in a falling state.

At this time, the overturned state detection unit 30 may further refer to the shape of the human region R1 in the image. In this case, for example, in addition to the above conditions, if the shape of the person region R1 in each image extends parallel to the floor surface, the falling state detection unit 30 determines that the person B1 is in a falling state. judge.

When the overturn detection unit 30 detects that the overturn of the person B1 has occurred, the overturn detection unit 30 erroneously detects a fall detection signal to notify an external alarm device (not shown). It is transmitted to the suppression unit 40 . Note that the fall detection unit 30 continuously transmits the fall detection signal to the erroneous detection suppression unit 40 while the falling state of the person B1 is being detected in the images sequentially acquired by the image acquisition unit 10. do.

Note that any method may be used by the falling state detection unit 30 to detect the occurrence of the falling state of the person B1. For example, the falling state detection unit 30 may detect the occurrence of the falling state of the person B1 directly from time-series images using a learned recursive neural network.

[False detection suppressor]
The erroneous detection suppression unit 40 is configured so that when the fall state detection unit 30 detects that the person B1 is falling (that is, when the fall state detection unit 30 issues a fall detection signal), the fall state detection unit 30 Check whether it is detected correctly or falsely detected.

As described above, in this type of image processing apparatus U, when the person B1 enters the shadow of an obstacle (here, the door B2) from the viewpoint of the imaging device 200, the person B1 appears as an obstacle in the image. Since the images are displayed integrally, a state in which the person B1 and the obstacle cannot be identified may occur. At this time, when using a method of detecting the occurrence of the overturned state of B1 from the temporal change of the images based on the time-series images (for example, the image difference method), the person B1 is hidden behind the obstacle. A situation in which there is no change in the image (for example, the person moved to the other side of the door) is a situation in which the person B1 falls and the image does not change, that is, a situation in which the amount of change per unit time becomes small. There is a risk of erroneous recognition. The erroneous detection suppression unit 40 suppresses erroneous detection caused by such a state.

FIG. 5 is a diagram showing an example of a monitoring target area to be determined for erroneous detection suppression by the erroneous detection suppression unit 40 .

In FIG. 5, Ball is an imaging area (that is, the entire area shown in the image captured by the imaging device 200), Bout is an area outside the imaging area Ball (here, a toilet area adjacent to the room), and B2 is an image of the person B1. A door for exiting from the area Ball to Bout outside the imaging area, B3 represents a monitoring target area.

Here, in the erroneous detection suppression unit 40, the area in front of the door B2 (ie, the area adjacent to the door B2) is set as the monitoring target area B3 (hereinafter also referred to as the "door front area B3"). This is because, as described above, when the person B1 leaves the imaging area Ball to the outside of the imaging area Bout through the door B2, there is a possibility that the overturned state detection unit 30 may make an erroneous detection.

6 and 7 are diagrams for explaining an example of determination processing of the erroneous detection suppression unit 40. FIG.

6 and 7 both show time-series state transitions of the imaging region Ball when the person B1 leaves the imaging region Ball to outside the imaging region Bout. Here, FIG. 6 shows time-series state transitions of the imaging region Ball when the fall state of the person B1 detected by the fall state detection unit 30 is not an erroneous detection, and FIG. The state transition of the imaging area Ball is shown when the detected falling state of the person B1 is erroneously detected.

Note that the timing T3 in FIGS. 6 and 7 is the timing at which the falling state detection unit 30 detects the falling state of the person B1. Timing T1 is timing 15 seconds before timing T3. Timing T4 is the timing after three seconds from the timing at which the overturning state detection unit 30 detects the overturning state of the person B1. Timing T2 is a timing between timing T1 and timing T3.

Specifically, the erroneous detection suppressing unit 40 selects the first image ( hereinafter referred to as an “after image”) (for example, an image at timing T4), and a person entering the monitoring target region B3 before the fall state detection unit 30 detects the occurrence of the fall state of the person B1. A second image (hereinafter referred to as a “before image”) (for example, an image at timing T1) is compared with the second image at the time when the image has not been taken. Then, the erroneous detection suppression unit 40 determines that the fall detection signal is erroneously detected when the amount of change between the After image and the Before image is small, and determines that the fall detection signal is erroneously detected when the amount of change is large. determine that it is not.

When the erroneous detection suppressing unit 40 determines that the fall detection signal is not erroneously detected, the erroneous detection signal is transmitted to an external alarm device (not shown) as it is so that the fall detection signal is not erroneously detected. If it is determined that there is, the transmission of the fall detection signal to an external alarm device (not shown) is suppressed.

Normally, when the person B1 falls, the person B1 appears in the area B3 in front of the door on the image (for example, the image at the timing T4 in FIG. 6). When the state does not occur, the person B1 has left the imaging area Ball to the outside of the imaging area Bout. 7) in the state of timing T4). The erroneous detection suppressing unit 40 performs the above determination processing using such basic image features.

Here, the “Before image” referred to by the false detection suppressing unit 40 is an image in which the person B1 does not exist in the door front area B3. This is an image taken at a timing before a predetermined time (for example, 15 seconds before) from the detection start timing at which the state detection unit 30 detects the occurrence of the falling state of the person B1. At this time, the "predetermined time" that goes back from the detection start timing can be appropriately changed depending on the size of the door front region B3 set as a comparison target, the motion speed of a person as a reference, and the like.

However, more preferably, the erroneous detection suppressing unit 40 detects the door front region by the human region detecting unit 20 among the time-series images acquired by the image acquiring unit 10 based on the detection result of the human region detecting unit 20 . An image when no person is detected in B3 is selected as a "Before image". As a result, it is possible to reliably select an image in which no person is present in the door front area B3 as the Before image.

The "after image" referred to by the erroneous detection suppressing unit 40 is, for example, the time-series image obtained by the image obtaining unit 10, which is obtained when the falling state of the person B1 is detected by the falling state detection unit 30. It is an image.

However, more preferably, the erroneous detection suppressing unit 40 is performed while the overturning state detection unit 30 detects the occurrence of the overturning state in the time-series images acquired by the image acquisition unit 10, and An image after a predetermined continuation determination time (for example, after 3 seconds) from the detection start timing is selected as an "after image". As a result, when the person B1 does not actually fall down, the image at the time before the person B1 finishes leaving the imaging region Ball to the outside of the imaging region Bout is suppressed from being selected as the After image. can do.

FIG. 8 is a diagram illustrating an example of a method by which the erroneous detection suppressing unit 40 obtains the amount of change in the After image with respect to the Before image.

For example, the erroneous detection suppression unit 40 subtracts the after image from the before image, and generates a histogram of pixel value differences between the before image and the after image as shown in FIG. In the histogram of FIG. 8, the horizontal axis represents the difference in pixel values (for example, pixel values of 256 gradations) between the Before image and the After image, and the vertical axis represents the appearance frequency of the difference (ie, the number of pixels).

Then, the erroneous detection suppressing unit 40 determines whether or not the number of pixels whose difference in pixel value is larger than a threshold value ThP (for example, 15) is less than a predetermined number (for example, 20) on the histogram. It is determined whether or not the falling state of the person B1 detected by 30 is an erroneous detection. That is, if the number of pixels in the histogram where the difference between the Before image and the After image is larger than the threshold value ThP is equal to or greater than a predetermined number, the erroneous detection suppressing unit 40 determines that the person B1 does not exist in both the Before image and the After image. Therefore, it is determined that the falling state of the person B1 is an erroneous detection. On the other hand, the erroneous detection suppressing unit 40 can specify that the person B1 appears only in the after image if the number of pixels on the histogram where the difference between the before image and the after image is larger than the threshold value ThP is a predetermined number or more. , it is determined that the overturned state of the person B1 is not an erroneous detection.

However, any other method may be used as the method for obtaining the amount of change between the Before image and the After image by the erroneous detection suppressing unit 40 . For example, the erroneous detection suppression unit 40 performs filter processing (for example, Laplacian filter) on each of the Before image and the After image in order to remove unnecessary noise components, and then obtains the amount of change between the Before image and the After image. good too.

FIG. 9 is a flow chart showing an example of the operation of the erroneous detection suppression unit 40. As shown in FIG.

In step S1, the erroneous detection suppressing unit 40 waits for detection of the falling state of the person B1 by the falling state detection unit 30 (S1: NO). S1: YES), the process proceeds to step S2.

In step S2, the erroneous detection suppression unit 40 generates an image of the door front area B3 a predetermined time (here, 15 seconds before) from the detection start timing at which the falling state of the person B1 is detected by the falling state detection unit 30. is stored in the RAM 103 as a Before image.

In step S3, the erroneous detection suppressing unit 40 waits for the state in which the overturning state detection unit 30 has issued the overturning detection signal to continue for the predetermined continuation determination time (S3: NO), and continues for the predetermined continuation determination time. If so (S3: YES), the process proceeds to step S4.

In step S4, the erroneous detection suppression unit 40 stores the image of the door front area B3 after the continuation determination time from the detection start timing in the RAM 103 as an after image.

In step S5, the erroneous detection suppression unit 40 generates a difference image between the After image and the Before image.

In step S6, the erroneous detection suppression unit 40 determines whether or not the difference between the after image and the before image is small based on the difference image between the after image and the before image. At this time, the erroneous detection suppression unit 40 generates, for example, a histogram of pixel value differences between the Before image and the After image as shown in FIG. It is determined whether or not the number of large pixels is less than a predetermined number (for example, 20). Then, if the difference between the After image and the Before image is small (S6: YES), the erroneous detection suppression unit 40 proceeds to step S7, and if the difference between the After image and the Before image is large (S6: NO), the process proceeds to Step S8. proceed to

In step S7, the erroneous detection suppression section 40 determines that the fall detection signal issued from the fall state detection section 30 is an erroneous detection, and cancels the fall detection signal.

In step S8, the erroneous detection suppression unit 40 determines that the fall detection signal issued from the fall state detection unit 30 is not an erroneous detection, and issues the fall detection signal to an external alarm device. Note that the external alarm device notifies that the person B1 has fallen, for example, through a speaker or a display in response to receiving the fall detection signal.

[effect]
As described above, according to the image processing apparatus U according to the present embodiment, it is possible to suppress erroneous detection when the person B1 falls down using a simple method.

(Other embodiments)
The present invention is not limited to the above-described embodiment, and various modifications are conceivable.

In the above-described embodiment, as an example of the configuration of the image processing apparatus 100, a mode in which the image processing apparatus 100 is arranged separately from the imaging apparatus 200 and acquires an image from the imaging apparatus 200 via the communication network 300 has been described. However, the image processing device 100 of the present invention may be arranged integrally with the imaging device 200 . In that case, the image processing apparatus 100 can be configured to acquire an image directly from the imaging apparatus 200 .

In the above embodiment, as an example of the configuration of the image processing apparatus 100, the functions of the image acquisition unit 10, the human area detection unit 20, the falling state detection unit 30, and the false detection suppression unit 40 are realized by one computer. Although described, it may of course be implemented by multiple computers. Programs and data read by the computer may also be distributed and stored in a plurality of computers.

Further, in the above embodiment, as an example of the operation of the image processing apparatus 100, the processes of the image acquisition unit 10, the human area detection unit 20, the falling state detection unit 30, and the false detection suppression unit 40 are executed in a series of flows. Although shown as being executed in parallel, part or all of these processes may of course be executed in parallel.

Further, in the above embodiment, the person B1 is shown as an application target of the image processing apparatus 100 . However, the image processing apparatus 100 of the present invention can also be applied to determine the overturned state of animals other than the person B1.

Although specific examples of the present invention have been described in detail above, these are merely examples and do not limit the scope of the claims. The technology described in the claims includes various modifications and changes of the specific examples illustrated above.

According to the image processing device according to the present disclosure, it is possible to suppress erroneous detection related to the occurrence of a falling state of a person using a simple method.

U Action Recognition System 10 Image Acquisition Unit 20 Human Area Detector 30 Overturned State Detector 40 False Detection Suppression Unit 100 Image Processing Device 200 Imaging Device 300 Communication Network Ball Imaging Area B1 Person B2 Door B3 Monitored Area Bout Area Outside Imaging Area

Claims (12)

an image acquisition unit that acquires time-series images generated by an imaging device that captures an image of a monitoring target area;
a falling state detection unit that detects the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected by the overturned state detection unit, a first image at the time when the occurrence of the overturned state is detected among the time-series images, and A second image at a time point a predetermined time before the detection start timing at which the occurrence of the overturned state is detected is extracted, and the overturned state detection unit detects the overturned state based on the amount of change in the first image with respect to the second image. an erroneous detection suppression unit that determines whether the result is an erroneous detection;
An image processing device comprising: an image acquisition unit that acquires time-series images generated by an imaging device that captures an image of a monitoring target area;
a falling state detection unit that detects the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected by the overturned state detection unit, the occurrence of the overturned state is detected by the overturned state detection unit in the time-series images, and A first image after a predetermined continuation determination time from the detection start timing at which the occurrence of the overturning state is detected and a second image at a time point before the occurrence of the overturning state is detected are extracted, and the second image is extracted. an erroneous detection suppression unit that determines whether or not the detection result of the overturned state detection unit is an erroneous detection, based on the amount of change in the first image with respect to the image;
An image processing device comprising: The erroneous detection suppressing unit determines that the detection result of the overturning state detecting unit is erroneous detection when the amount of change in the first image with respect to the second image is less than a predetermined value. If the amount of change in the first image relative to the
The image processing apparatus according to claim 1 or 2 . The monitoring target area is an area adjacent to a door for the target person to exit from the imaging area to the outside of the imaging area, out of the entire imaging area imaged by the imaging device.
The image processing apparatus according to any one of claims 1 to 3 . Further comprising a human region detection unit that detects whether or not the target person exists in the monitoring target region based on the time-series images,
The image processing apparatus according to any one of claims 1 to 4 . The erroneous detection suppressing unit selects, from among the time-series images, an image at a time when the target person is not detected in the monitoring target area as the second image.
The image processing apparatus according to claim 5 . The erroneous detection suppressing unit selects, from among the time-series images, an image at a point in time a predetermined time before the detection start timing at which the occurrence of the overturned state is detected by the overturned state detection unit, as the second image. do,
An image processing apparatus according to claim 2 and any one of claims 3 to 6 depending on claim 2 . The erroneous detection suppression unit controls detection start timing at which the occurrence of the overturning state is detected while the occurrence of the overturning state is being detected by the overturning state detection unit in the time-series images. selecting an image after a predetermined continuation determination time from as the first image;
7. An image processing apparatus according to claim 1 and any one of claims 3 to 6 depending on claim 1 . Acquiring time-series images generated by an imaging device that captures an image of a monitored area,
detecting the occurrence of a falling state of a subject existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the first image at the time when the occurrence of the overturned state is detected among the time-series images, and from the detection start timing at which the occurrence of the overturned state is detected. , extracting a second image at a point in time a predetermined time ago , and determining whether or not the occurrence of the falling state is an erroneous detection based on the amount of change in the first image with respect to the second image;
A control method for an image processing device. Acquiring time-series images generated by an imaging device that captures an image of a monitored area,
detecting the occurrence of a falling state of a subject existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the period during which the occurrence of the overturned state is detected in the time-series images and from the detection start timing at which the occurrence of the overturned state is detected extracting a first image after a predetermined continuation determination time and a second image at a point in time before the occurrence of the overturned state is detected, and based on a change of the first image with respect to the second image; , determining whether the occurrence of the falling state is an erroneous detection;
A control method for an image processing device. to the computer,
A process of acquiring time-series images generated by an imaging device that captures an image of a monitoring target area;
a process of detecting the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the first image at the time when the occurrence of the overturned state is detected among the time-series images, and from the detection start timing at which the occurrence of the overturned state is detected. , a process of extracting a second image at a point in time a predetermined time ago , and determining whether or not the occurrence of the overturned state is an erroneous detection based on the amount of change of the first image with respect to the second image;
A control program for an image processing apparatus that executes to the computer,
A process of acquiring time-series images generated by an imaging device that captures an image of a monitoring target area;
a process of detecting the occurrence of a falling state of a target person existing in the monitoring target area based on the time-series images;
When the occurrence of the overturned state is detected, the period during which the occurrence of the overturned state is detected in the time-series images and from the detection start timing at which the occurrence of the overturned state is detected extracting a first image after a predetermined continuation determination time and a second image at a point in time before the occurrence of the overturned state is detected, and based on a change of the first image with respect to the second image; , a process of determining whether the occurrence of the falling state is an erroneous detection;
A control program for an image processing apparatus that executes

Images