JP7176719B2 - DETECTION DEVICE, DETECTION SYSTEM, DETECTION METHOD AND PROGRAM - Google Patents

Description

The present invention relates to a detection device, detection system, detection method and program.

When transporting using a transport device such as a belt conveyor, transported objects may fall and accumulate during transport. In that case, a person may find the sediment by visual inspection and clean it. However, in order to improve work efficiency, it is desirable to detect deposits automatically. As one of methods for automatically detecting deposits, a method of monitoring with a camera and detecting an object to be detected (that is, deposits) from a photographed image is conceivable. Patent Documents 1 and 2 describe techniques for detecting a detection target from a captured image.

In the technique described in Patent Document 1, a plurality of dictionary information indicating characteristics of a detection target (that is, a person) is stored, and a person is detected based on a captured image and the plurality of dictionary information. technology is described.

Patent Literature 2 describes a technique for detecting a detection target by dividing one detection target into a plurality of templates and performing template matching. Specifically, in the technique described in Patent Document 2, when a pedestrian is a detection target, a template corresponding to the upper part of the pedestrian (including the head and shoulders) and a middle part of the pedestrian ( A template corresponding to the lower part of the pedestrian (including the waist) and a template corresponding to the lower part of the pedestrian (including the leg) are prepared in advance. Then, in the technique described in Patent Document 2, a detection target (that is, a pedestrian) is detected based on the matching rate of each template.

JP 2016-171526 A JP 2010-128919 A

It should be noted that the disclosure of the prior art document mentioned above is incorporated herein by reference. The following analysis is made in light of the present invention.

As described above, it is desirable to automatically detect deposits in order to improve work efficiency. Here, in the technique described in Patent Document 1, it is required to be able to presume the pattern of the state of the object to be detected (that is, the person). Further, in the technique described in Patent Document 2, it is necessary that the ratio of the height and width of the object to be detected is within a predetermined range. However, it is difficult to predict in advance the shape and size of the deposit if the object to be transported falls during transportation. Therefore, it is difficult to detect deposits from captured images using the techniques described in Patent Documents 1 and 2.

SUMMARY OF THE INVENTION Accordingly, it is an object of the present invention to provide a detection apparatus, a detection system, a detection method, and a program that contribute to easy detection of a detection target while flexibly coping with changes in the shape of the detection target.

According to a first aspect, a detection device is provided. The detection device includes a model storage unit that stores, as a prediction model, a feature amount of a local region of the texture of the detection target.
Further, the detection device acquires a captured image, cuts out a child region of a predetermined window size from the captured image, and determines the presence or absence of the detection target based on the child region and the prediction model. A detection control unit is provided.

According to a second aspect, a detection system is provided. The detection system includes a monitoring device having a camera that can pan and tilt, and a detection device that determines the presence or absence of an object to be detected.
The detection device includes a model storage unit that stores, as a prediction model, a feature amount of a texture of a local area among the textures of the detection object.
Further, the detection device acquires the photographed image from the monitoring device, cuts out a child region of a predetermined window size from the photographed image, and determines the detection target object based on the child region and the prediction model. A detection control unit is provided to determine the presence or absence of the device.

According to a third aspect, a detection method is provided. The detection method includes a step of storing, by a computer having a storage unit, a feature amount of texture of a local region among textures of a detection target in the storage unit as a prediction model.
Furthermore, the detection method includes a step of acquiring a photographed image by the computer.
Further, the detection method includes a step of cutting out a child area of a predetermined window size from the captured image by the computer.
Further, the detection method includes the step of determining the presence or absence of the detection target based on the child region and the prediction model.
It should be noted that this method is tied to a specific machine, a detection device that determines the presence or absence of an object to be detected.

A fourth aspect provides a program. The program causes a computer having a storage unit to execute a process of storing, in the storage unit, the feature amount of the texture of the local region among the textures of the detection object as a prediction model.
Furthermore, the program causes the computer to execute a process of acquiring a captured image.
Furthermore, the program causes the computer to execute a process of cutting out a child area of a predetermined window size from the captured image.
Furthermore, the program causes the computer to execute a process of determining the presence or absence of the detection target based on the child region and the prediction model.
These programs can be recorded in a computer-readable storage medium. The storage medium can be non-transient such as semiconductor memory, hard disk, magnetic recording medium, optical recording medium, and the like. The invention can also be embodied as a computer program product.

ADVANTAGE OF THE INVENTION According to this invention, the detection apparatus, the detection system, the detection method, and the program which contribute to detecting a detection target easily, while responding flexibly to the shape change of a detection target are provided.

1 is a diagram for explaining an overview of an embodiment; FIG. 1 is a block diagram showing an example of the overall configuration of a detection system 100; FIG. FIG. 4 is a diagram showing an example of an area that can be photographed by the cameras 401 to 403 when the cameras 401 to 403 are not rotated; FIG. 4 is a diagram showing an example of an area that can be photographed by the camera 11 when the camera 11 is rotated; 4 is a flow chart showing an example of the operation of the detection system 100; It is a figure which shows an example of the texture of a local area|region, and a detection result. It is a figure which shows an example of the problem of the background subtraction method. 3 is a block diagram showing an example of hardware of the detection device 20. FIG.

First, an overview of one embodiment will be described with reference to FIG. It should be noted that the drawing reference numerals added to this outline are added to each element for convenience as an example to aid understanding, and the description of this outline does not intend any limitation. Also, connecting lines between blocks in each block diagram include both bi-directional and uni-directional. The unidirectional arrows schematically show the flow of main signals (data) and do not exclude bidirectionality. Furthermore, in the circuit diagrams, block diagrams, internal configuration diagrams, connection diagrams, etc. disclosed in the present application, an input port and an output port exist at the input end and the output end of each connection line, respectively, although not explicitly shown. Input/output interfaces are similar.

As described above, there is a demand for a detection device that contributes to easy detection of a detection target while flexibly coping with changes in the shape of the detection target.

Therefore, as an example, a detection device 1000 shown in FIG. 1 is provided. The detection device 1000 includes a model storage section 1001 and a detection control section 1002 . A model storage unit 1001 stores, as a prediction model, a feature amount of a local region of the texture of the detection target. The detection control unit 1002 acquires a captured image and cuts out a child area of a predetermined window size from the captured image. Then, the detection control unit 1002 compares the child region extracted from the captured image with the prediction model, and determines the presence or absence of the detection target.

Therefore, the detection apparatus 1000 determines the presence or absence of the detection target without using information regarding the shape and size of the detection target. Therefore, even if the shape and size of the detection target cannot be assumed in advance, the detection apparatus 1000 can detect the presence or absence of the detection target by using the texture of the local area among the textures of the detection target. Easy to judge. Therefore, the detection apparatus 1000 contributes to easy detection of a detection target while flexibly coping with changes in the shape of the detection target.

[First Embodiment]
A first embodiment will be described in detail with reference to the drawings. In the following description, deposits will be exemplified as detection targets. However, this does not mean that the object to be detected in the detection system 100 according to the present embodiment is limited to deposits.

FIG. 2 is a block diagram showing an example of the overall configuration of the detection system 100 according to this embodiment. The detection system 100 includes a monitoring device 10 , a detection device 20 and a terminal 30 . 2, one monitoring device 10 and one terminal 30 are shown. However, this does not mean that the detection system 100 is limited to including one monitoring device 10 and one terminal 30, respectively. 2 is an example of the configuration of the monitoring device 10, the detection device 20, and the terminal 30. As shown in FIG. Moreover, FIG. 2 is not meant to limit the configurations of the monitoring device 10, the detection device 20, and the terminal 30 to the configurations shown in FIG.

The monitoring device 10 is an information processing device (computer) that captures an image of a monitored area and transmits the captured image to the detection device 20 . The monitoring device 10 is provided with a camera 11 capable of panning (horizontal swing), tilting (vertical swing), and zooming. The monitoring device 10 rotates the camera 11 within the movable range and transmits a plurality of captured images captured by the camera 11 to the detecting device 20 . In addition, the monitoring device 10 changes the zoom ratio of the camera 11 within the movable range, and transmits a plurality of captured images captured by the camera 11 to the detection device 20 .

The detection device 20 is an information processing device (computer) that determines the presence or absence of a detection target (for example, deposits) from a captured image. The detection device 20 connects to the monitoring device 10 and the terminal 30 via a network.

The terminal 30 is an information processing device (computer) that receives and outputs the presence or absence of deposits, the position of deposits, captured images, and the like from the detection device 20 .

[Configuration of monitoring device]
Next, the monitoring device 10 will be described in detail. The monitoring device 10 includes a camera 11 , a monitoring device communication section 12 and a monitoring control section 13 .

Camera 11 takes an image. As described above, the camera 11 is assumed to be capable of so-called panning and tilting. Furthermore, it is assumed that the camera 11 can change the zoom ratio.

FIG. 3 is a diagram showing an example of an area that can be photographed by the cameras 401 to 403 when the cameras 401 to 403 are not rotated around the vertical axis and around the horizontal axis. Assume that deposits B11, B12, and B13 are present in regions R11, R12, and R13 surrounded by dashed lines. Suppose that areas R11, R12, and R13 are areas that can be photographed by the cameras 401, 402, and 403, respectively. In that case, it is necessary to install a plurality of cameras (cameras 401 to 403) in order to photograph the deposits B11, B12, and B13.

On the other hand, FIG. 4 is a diagram showing an example of an area that can be photographed by the camera 11 when the camera 11 is rotatable (so-called panning and tilting is possible) around the vertical axis and around the horizontal axis. Assume that deposits B21, B22, and B23 are present in a region R21 surrounded by a dashed line. In that case, by installing one camera 11, the deposits B21, B22, and B23 can be photographed. Therefore, by rotating the camera 11, it is possible to photograph a wider range than when the cameras 401 to 403 are not rotated (FIG. 3).

The monitoring device communication unit 12 controls communication with the detection device 20 .

The monitoring control unit 13 controls the camera 11 . For example, the monitoring control unit 13 controls the rotation angle and zoom ratio of the camera 11 to control the imaging area.

The monitoring control unit 13 can rotate the camera 11 around the vertical axis. Similarly, the monitoring control unit 13 can rotate the camera 11 around the horizontal axis. The monitoring control unit 13 causes the camera 11 to generate a photographed image each time the angle of the camera 11 is changed (that is, each time the photographing area is changed). Furthermore, the monitoring control unit 13 may change the zoom ratio of the camera 11 . The monitoring control unit 13 may generate a captured image on the camera 11 each time the zoom ratio of the camera 11 is changed. After generating the captured image, the monitoring control unit 13 transmits the generated captured image to the detection device 20 via the monitoring device communication unit 12 .

Therefore, the monitoring device 10 generates a captured image of the space within the movable range of the camera 11 and transmits the generated captured image to the detection device 20 . As a result, the detection device 20 can acquire captured images over a wider range than when the camera 11 is not rotated.

The monitoring control unit 13 swings the camera 11 up, down, left, and right (that is, pans and tilts the camera 11), and causes the camera 11 to generate a photographed image of the space within the movable range of the camera 11. FIG. Then, the monitoring control unit 13 transmits the generated multiple (two or more) captured images to the detection device 20 . The monitoring control unit 13 causes the camera 11 to generate captured images regarding the space within the movable range of the camera 11, and performs a process of transmitting a plurality of (two or more) generated captured images to the detection device 20 at predetermined time intervals. repeat to In the following description, a series of processing for swinging the camera 11 up, down, left, and right (that is, panning and tilting the camera 11) and causing the camera 11 to generate a photographed image with respect to the space within the movable range of the camera 11 is referred to as "one This is referred to as a single shooting process.

[Configuration of detection device]
Next, the detection device 20 will be described in detail. The detection device 20 includes a detection device communication section 21 , an image storage section 22 , a model storage section 23 and a detection control section 24 .

The detection device communication unit 21 controls communication with the monitoring device 10 and the terminal 30 . When the detection device communication unit 21 receives the captured image from the monitoring device 10 , the detection control unit 24 stores the captured image received by the detection device communication unit 21 in the image storage unit 22 .

The image storage unit 22 stores captured images generated (captured) by the camera 11 .

The model storage unit 23 stores, as a prediction model, the feature amount of the local area among the deposit textures. The local area texture may be a periodic pattern texture. The feature amount may be information related to texture color, brightness, pattern, and the like.

Here, it is assumed that the prediction model includes a feature amount generated using machine learning with respect to the texture of the local region among the textures of the deposits. For example, the prediction model may be a local region texture feature learned using deep learning such as a neural network.

The detection control unit 24 acquires a captured image, and cuts out a child area having a predetermined window size from the captured image. Then, the detection control unit 24 determines the presence or absence of deposits based on the extracted child region and the prediction model stored in advance in the model storage unit 23 . Specifically, the detection control unit 24 determines the presence or absence of deposits based on whether a child region corresponding to the texture of the deposits exists in the captured image.

Then, the detection control unit 24 extracts a plurality of child regions from a predetermined range of the captured image, performs collation (matching) with respect to the plurality of child regions with the prediction model, and outputs the result of matching with the prediction model. Output. The detection control unit 24 may transmit the result of matching between the clipped child region and the prediction model to the terminal 30 used by the user.

Appropriate local region textures differ depending on the shooting environment, the number of cameras (the number of monitored locations), detection targets (deposits), and the like. Here, the “appropriate local area texture” means a texture including necessary and sufficient feature amounts for determining the presence or absence of deposits.

Therefore, in the detection system 100 according to this embodiment, the window size may be determined based on the user's operation. The user may adjust the window size while viewing the output of the detection device 20 (result of matching with the prediction model). For example, the window size may be determined by the user adjusting the window size based on a predetermined percentage (for example, 1%) of the window size with respect to the entire deposition area in the captured image. In other words, the detection control unit 24 may update the window size based on the user's operation.

In addition, if a fence, a pillar, or the like is installed in front of the camera 11, and deposits exist behind the fence, the pillar, or the like, part of the deposit is hidden by the fence, the pillar, or the like (that is, occlusion may occur). However, in the detection system 100 according to the present embodiment, the user can adjust the window size so that the window size is smaller than the deposits that can be seen through gaps between fences, pillars, and the like, thereby avoiding the effects of occlusion. can be used to determine the presence or absence of deposits.

[Terminal configuration]
Next, the terminal 30 will be described in detail. The terminal 30 includes a terminal communication section 31 , a terminal control section 32 and an output section 33 .

The terminal communication unit 31 controls communication with the detection device 20 . When the terminal communication unit 31 receives detection results (presence or absence of deposits, positions of detected deposits), photographed images, etc. from the detection device 20, the terminal control unit 32 receives the received detection results, photographed images, etc. , to the output unit 33 . The output unit 33 outputs detection results (presence or absence of deposits, positions of detected deposits), photographed images, and the like to a file or screen.

[Detection system operation]
The operation of detection system 100 will now be described in detail.

FIG. 5 is a flow chart showing an example of the operation of the detection system 100 according to this embodiment.

In step S<b>1 , the detection control unit 24 reads a prediction model from the model storage unit 23 .

When the captured image is acquired (step S2), the detection control unit 24 cuts out a child area having a predetermined window size from the acquired captured image (step S3). Then, the detection control unit 24 determines whether or not the extracted sub-region includes the texture of the deposit (step S4).

If the extracted child region does not include the texture of the deposit (No branch of step S4), the process transitions to step S6. On the other hand, when the texture of the deposit is included in the extracted child region (Yes branch of step S4), the detection control unit 24 saves the detection result (step S5).

FIG. 6 is a diagram illustrating an example of textures of local regions and detection results. FIG. 6(a) is a diagram showing an example of the texture of the local area. Rectangular regions w1 to w3 shown in FIG. 6(a) are examples of local regions corresponding to prediction models. FIG.6(b) is a figure which shows an example of a detection result.

For example, it is assumed that the model storage unit 23 stores in advance feature amounts relating to the textures of the rectangular regions w1 to w3 as prediction models. In that case, when the detection device 20 acquires the captured image to be determined, the detection device 20 compares the child regions cut out from the captured image with the prediction model (features related to the texture of the rectangular regions w1 to w3), and compares the extracted child regions. Determine whether the region contains the texture of the predictive model. Then, the detection device 20 transmits the detection result (for example, the image shown in FIG. 6B) to the terminal 30 .

In step S6, the detection control unit 24 determines whether or not the entire captured image has been scanned. That is, the detection control unit 24 determines whether or not a child region is cut out from the entire photographed image and the presence or absence of deposit texture is determined.

If the entire captured image has not been scanned (No branch in step S6), the process returns to step S3 to continue the process. In other words, the detection control unit 24 cuts out a child area of a predetermined window size from the unscanned area of the captured image. Then, the detection control unit 24 performs a process of determining whether or not the texture of the deposit is included in the newly cut child area. In this manner, it is determined whether or not the photographed image contains the texture of the deposit.

When the entire captured image is scanned (Yes branch of step S6), the detection control unit 24 notifies the terminal 30 of the detection result (step S7).

The terminal 30 performs processing such as displaying the detection result on the display, and presents the detection result to the user. The user may adjust (change) the window size by viewing the presented detection results. Specifically, the user may input a new window size to terminal 30 or detection device 20 . When the user inputs a new window size to the terminal 30, the terminal 30 naturally notifies the detection device 20 of the input window size.

Next, with reference to FIG. 7, differences between the case of detecting deposits using the background subtraction method and the detection system 100 according to the present embodiment will be described. For example, assume that an image BG31 shown in FIG. 7A is prepared in advance as a background image. It is assumed that a region R31 surrounded by a dashed line shown in FIG. 7(a) includes a plant.

Assume that the camera 11 captures the image IMG32 shown in FIG. 7B after capturing the image BG31. In that case, a device (computer) capable of executing image processing can detect the deposit B31 included in the image IMG32 by taking the difference between the image BG31 and the image IMG32.

However, suppose that the camera 11 captures the image IMG33 shown in FIG. 7C after capturing the image BG31. Here, in image IMG33, it is assumed that a region R32 surrounded by a dashed line includes a plant that has grown since BG31 was captured. In other words, as the plant grows, the appearance (shape, size, color, etc.) is different from when the BG 31 was photographed. In that case, a device (computer) capable of executing image processing detects the plant and deposit B32 included in region R32 by taking the difference between image BG31 and image IMG33. Therefore, if there is a possibility that an object different from the object to be detected has a different appearance than when the background image was captured, using the background subtraction method may result in detecting the object that is different from the object to be detected. have a nature.

On the other hand, in the detection system 100 according to the present embodiment, the detection device 20 determines the presence/absence of the detection target in the captured image based on the texture of the local area among the textures of the detection target. Therefore, in the detection system 100 according to the present embodiment, even when the appearance of an object different from the detection target changes, the detection target can be detected appropriately.

As described above, the detection system 100 according to the present embodiment uses the texture of the local area among the textures of the detection target. Accordingly, in the detection system 100 according to the present embodiment, the detection device 20 can appropriately detect the detection target even when the shape, size, etc. of the detection target in the captured image change. Therefore, the detection system 100 according to the present embodiment contributes to easy detection of a detection target while flexibly coping with changes in the shape of the detection target.

In the detection system 100 according to this embodiment, by using the texture of the local area among the textures of the detection target, the detection target can be detected while moving the camera 11 . Therefore, the detection system 100 according to this embodiment contributes to appropriately and easily detecting a detection target while reducing the number of cameras.

In addition, the detection system 100 according to the present embodiment uses the texture of the local area among the textures of the detection target to suppress the influence of the surrounding environment (growing of plants, changes in brightness due to weather, etc.). contribute to

Further, in the detection system 100 according to the present embodiment, by using the texture of the local region among the textures of the detection target, the detection device 20 can detect the Detection target can be detected. Therefore, the detection system 100 according to this embodiment contributes to appropriately detecting the detection target while improving the degree of freedom of the monitoring position (installation position of the camera 11).

Further, in the detection system 100 according to the present embodiment, the number of images required to generate a model (prediction model) is reduced by using the feature amount of the texture of the local region among the textures of the detection target. contribute to reducing

Next, hardware of the detection device 20 will be described.

FIG. 8 is a block diagram showing an example of hardware of the detection device 20. As shown in FIG. The detection device 20 can be configured by a computer and has the configuration illustrated in FIG. For example, the detection device 20 includes a CPU (Central Processing Unit) 1, a communication interface 2, a memory 3, an auxiliary storage device 4, etc., which are interconnected by an internal bus.

The functions of the detection device 20 are realized by the CPU 1 reading the captured images and prediction models stored in the auxiliary storage device 4 and executing the programs stored in the memory 3 . Note that the configuration of the monitoring device 10 and the terminal 30 can be the same as the hardware configuration of the detection device 20 shown in FIG. 8, so detailed description thereof will be omitted. Note that the monitoring device 10 can be realized by further including a camera 11 . Moreover, the terminal 30 can be realized by further including a display.

In the above description, the monitoring device 10, the detection device 20, and the terminal 30 are described as different devices. However, it goes without saying that the detecting device 20 may be configured to include the functions of the monitoring device 10 and the terminal 30 .

Moreover, in the above description, the prediction model has been illustrated and described including the feature amount generated using machine learning. However, this is not intended to limit the prediction model. It goes without saying that the prediction model may include features generated by a method other than machine learning.

In addition, the camera 11 may capture a moving image captured by rotating within the movable range or by changing the zoom ratio. In that case, the monitoring device 10 transmits the moving image captured by the camera 11 to the detecting device 20 . For example, the monitoring control unit 13 shakes the camera 11 up, down, left, and right (that is, pans and tilts the camera 11) at predetermined time intervals, and generates a moving image on the camera 11 regarding the space within the movable range of the camera 11. You can let me (take a picture). In this case, the detection control unit 24 acquires moving images captured by rotating the camera 11 within the movable range or changing the zoom ratio, and extracts still images from the acquired moving images as captured images. Then, the detection control unit 24 determines the presence or absence of deposits with respect to the plurality of photographed images extracted from the moving image.

Moreover, in the above description, a form in which the photographing area is changed by rotating the camera 11 has been described. However, this does not mean that the imaging area is limited by rotating the camera 11 .

Also, the monitoring device 10 may be mounted on a moving body (vehicle, movable robot, etc.). In that case, the monitoring device 10 may transmit the image captured by the camera 11 to the detection device 20 as the captured image while the mobile object is moving. In other words, the detection device 20 may change the deposit detection range by moving the moving object.

Further, for example, in the case of detecting a deposit generated by a transported object falling from a conveying device (such as a belt conveyor), the camera 11 may be installed at a position where the side of the conveying device can be photographed. In that case, the detection device 20 may determine the height of the deposit relative to the transport device by comparing the position of the transport device in the captured image with the position of the top of the detected deposit. For example, the user of the terminal 30 or the like may input the position of the transport device in the captured image to the detection device 20 in advance.

The monitoring device 10 may also include a range image sensor. The range image sensor may be a stereo camera, a ToF (Time of Flight) camera, or a Lidar (Light Detection and Ranging). Then, the monitoring device 10 may transmit the information (image) generated by the distance image sensor and the captured image generated by the camera 11 to the detection device 20 . In that case, the detection control unit 24 determines the presence or absence of deposits based on the captured image generated by the camera 11 . The detection control unit 24 measures the height of the object to be detected based on the output of the distance image sensor when the child area corresponding to the texture of the deposit exists in the captured image. For example, when the detection device 20 detects the texture of the deposit in the photographed image, based on the information (image) generated by the distance image sensor, the position of the top of the detected deposit is determined as the height of the deposit. can be estimated as

It should be noted that the disclosure of the above-mentioned patent document is incorporated herein by reference. Within the framework of the full disclosure of the present invention (including the scope of claims), modifications and adjustments of the embodiments are possible based on the basic technical concept thereof. Also, within the framework of the full disclosure of the present invention, various combinations or selections (partial (including deletion) is possible. That is, the present invention naturally includes various variations and modifications that can be made by those skilled in the art according to the entire disclosure including claims and technical ideas. In particular, any numerical range recited herein should be construed as specifically recited for any numerical value or subrange within that range, even if not otherwise stated. Where the present invention presents algorithms, software, or flow charts or automated process steps, it is obvious that computers are used and that computers are associated with processors and memory or storage devices. is. Therefore, even in the absence of explicit indication, it is to be understood that these elements are described herein.

1 CPU
2 Communication interface 3 Memory 4 Auxiliary storage device 10 Monitoring device 11, 401 to 403 Camera 12 Monitoring device communication unit 13 Monitoring control unit 20, 1000 Detecting device 21 Detecting device communication unit 22 Image storage unit 23, 1001 Model storage unit 24, 1002 Detection control unit 30 Terminal 31 Terminal communication unit 32 Terminal control unit 33 Output unit 100 Detection system

Claims (9)

a model storage unit that stores, as a prediction model, the feature amount of the texture of the local area among the textures of the detection target;
Acquire the captured image,
Detection for extracting a child region of a predetermined window size from the captured image, comparing the child region with the prediction model, and determining that the detection target object exists when the child region includes the texture of the prediction model. and a detection control unit that repeats for the entire area of the captured image, storing the detection result in association with the child area,
The detection control unit transmits to a terminal used by a user, the detection result in which the child area determined to contain the detection target is identifiable from the other child areas, and responds to the user's operation. and updating the window size based on . 2. The detection device according to claim 1, wherein said prediction model includes a feature quantity generated using machine learning with respect to the texture of said local area among the textures of said detection object. 2. The detection control unit obtains a moving image captured by rotating the camera within a movable range or changing a zoom ratio, and extracts a still image from the moving image as the captured image. 3. The detection device according to 2 . a surveillance device comprising a camera that can be panned and tilted;
a detection device for determining the presence or absence of an object to be detected;
consists of
The detection device is
a model storage unit that stores, as a prediction model, a feature amount of a texture of a local area among the textures of the detection target;
Acquiring a captured image from the monitoring device,
Detection for extracting a child region of a predetermined window size from the captured image, comparing the child region with the prediction model, and determining that the detection target object exists when the child region includes the texture of the prediction model. and a detection control unit that repeats for the entire area of the captured image, storing the detection result in association with the child area,
The detection control unit transmits to a terminal used by a user the detection result in which the child area determined to contain the detection target is identifiable from other child areas, A detection system that updates the window size based on . 5. The detection system according to claim 4, wherein the monitoring device rotates the camera within a movable range and transmits the plurality of captured images captured by the camera to the detection device. Mounting the monitoring device on a moving body,
6. The detection system according to claim 4, wherein said monitoring device transmits an image captured by said camera as said captured image to said detection device while said mobile body is moving. The monitoring device further comprises a range image sensor,
The detection control unit measures the height of the detection target based on the output of the distance image sensor when the child region corresponding to the texture of the detection target exists in the captured image . Item 7. A detection system according to any one of Items 4 to 6 . A computer having a storage unit
a step of storing, as a prediction model, a feature amount of a texture of a local area among the textures of the object to be detected in the storage unit;
a step of acquiring a photographed image;
a step of cutting out a child region of a predetermined window size from the captured image;
The child region and the prediction model are collated, and if the child region includes the texture of the prediction model, detection is performed to determine that the detection object exists, and the detection result is stored in association with the child region. a step of repeating this for the entire area of the captured image;
a step of transmitting to a terminal used by a user the detection result in which the child area determined to contain the detection target is identifiable from the other child areas;
updating the window size based on the user's operation ;
detection methods, including A computer having a storage unit,
A process of storing, as a prediction model, the feature amount of the texture of the local area among the textures of the detection target in the storage unit;
a process of acquiring a photographed image;
A process of cutting out a child area of a predetermined window size from the captured image;
The child region and the prediction model are collated, and if the child region includes the texture of the prediction model, detection is performed to determine that the detection object exists, and the detection result is stored in association with the child region. A process of repeating this for the entire area of the captured image;
a step of transmitting to a terminal used by a user the detection result in which the child area determined to contain the detection target is identifiable from the other child areas;
a process of updating the window size based on the user's operation ;
program to run.

Images