JP6746112B2 - Monitoring device and monitoring method - Google Patents

Description

The present invention relates to a monitoring device and a monitoring method.

Conventionally, a monitoring device that acquires and monitors an image of a predetermined place has been proposed (Patent Document 1).

Japanese Patent No. 3652666

However, in the conventional technique, in order to confirm the abnormality, it is necessary for a person to look at the image to make a judgment, and it is necessary to see the image without abnormality. Further, when the place where the abnormality occurs is unknown, it is necessary to arrange a large number of monitoring devices, which increases the scale of the system.

The present invention has attempted to solve such a problem, and an object of the present invention is to provide a monitoring apparatus and a monitoring method in which images viewed by humans are minimal and the system scale is small.

A first aspect of the invention is an information receiving unit that receives search target information indicating a search target, an image acquiring unit that is arranged in a moving body and acquires an image of an object, and the object included in the image is the search target. Object recognition means for determining whether or not the position of the object, position information generation means for generating position information indicating the position of the object, and transmission information generation means for generating transmission information including the image of the object and the position information. Is a monitoring device having.

According to the configuration of the first invention, the monitoring device can generate transmission information including an image of an object determined to be a search target and position information. Therefore, it is sufficient for the person in charge of monitoring to check only the transmission information transmitted from the monitoring device, so that the image to be viewed can be minimized. Further, since the image acquisition means is arranged on the moving body, it is possible to acquire images at different positions by using one moving body. Therefore, the scale of the system of the monitoring device can be reduced.

A second invention is the monitoring device according to the first invention, wherein the object recognition means includes data generated by deep learning.

A third aspect of the present invention is the configuration of the first or second aspect of the present invention, further comprising moving state information generating means for generating moving state information indicating a moving state of the object, wherein the transmission information is the moving state. It is a monitoring device including information.

In a fourth aspect of the invention, in the configuration of the third aspect of the invention, based on the movement state information, a position prediction unit that generates predicted position information indicating a predicted position of the object after a predetermined time has elapsed, the transmission information. Is a monitoring device including the predicted position information.

A fifth aspect of the present invention is the monitoring device according to any one of the first to fourth aspects of the invention, wherein the object recognition means acquires the feature information indicating the feature of the search target by deep learning.

According to a sixth aspect of the invention, in the configuration of any one of the first to fifth aspects, at least one of the information receiving unit, the object recognizing unit, the position information generating unit, and the transmission information generating unit is It is a monitoring device included in a management device that manages a moving body.

A seventh invention is an information receiving step of receiving search target information indicating a search target,
An image acquisition step of acquiring an image by an image acquisition means for acquiring the image, which is arranged on the moving body,
An object recognition step of determining whether the object included in the image is the search target,
A position information generating step of generating position information indicating the position of the object,
A transmission information generating step of generating transmission information including the image of the object and the position information,
Is a monitoring method.

As described above, according to the present invention, the image viewed by a person is sufficient, and the system of the monitoring device is also small.

It is a figure which shows the outline of one Embodiment of this invention. It is a schematic block diagram of this embodiment. It is a figure which shows an example of the image of search object. It is a figure which shows an example of transmission information. It is a flowchart of this embodiment. It is a figure which shows an example of transmission information. It is a figure which shows an example of transmission information. It is a figure which shows the modification of the image of search object. 3 is a flowchart of an embodiment of the present invention. 3 is a flowchart of an embodiment of the present invention. It is a figure which shows the outline of one Embodiment of this invention. 3 is a flowchart of an embodiment of the present invention.

Embodiments of the present invention will be described with reference to the drawings. It should be noted that description of configurations that can be appropriately performed by those skilled in the art will be omitted, and only the basic configuration of the present invention will be described.

<First embodiment>
As shown in FIG. 1, the moving body 1 is arranged in a theme park having a ferris wheel 100, a castle 102, a fountain 104 and trees 106. The mission of the mobile unit 1 is to transmit the image and position of an object considered to be a search target to the base station 50 (see FIG. 2) or the client device 60 (see FIG. 2). The search target is a specific person or object, or a person or object having a specific property.

In the present embodiment, the search target of the mobile unit 1 is the person shown in the data 70 (see FIG. 3). The person shown in the data 70 is, for example, a person who is wanted as a criminal suspect. Then, the person shown in the data 70 is disguised as another person wearing sunglasses and disguised.

The appearance of the moving body 1 is configured as a character such as an animal so that it matches the atmosphere of the theme park and does not give a sense of caution to the person or the like who is the search target. The moving body 1 is an animal character, and at least one of both eyes is a camera 3. The camera 3 is a digital camera that captures a digital image, a video camera, or the like. The eyelashes of the mobile unit 1 are antennas 5 for transmitting and receiving radio waves so that they can communicate with the outside via the communication satellite 40 and receive positioning radio waves from a GPS (Global Position System) satellite. Has become. Further, the moving body 1 has four tires 7. The tire 7 is a moving unit that moves the moving body 1, and is connected to a motor inside the moving body 1. The moving body 1 is provided with a communication device, an autonomous movement control device, a GPS device, an inertial sensor such as an acceleration sensor and a gyro sensor, a barometer, and a battery. The mobile unit 1 can communicate with the base station 50 (see FIG. 2) and the client device 60 (see FIG. 2) by the communication device. The moving body 1 is an example of a moving body, and is also an example of a monitoring device. The mobile unit 1, the base station 50, and the client device 60 constitute a monitoring system as a whole.

As shown in FIG. 2, the mobile unit 1 includes a CPU (Central Processing Unit) 10, a storage unit 12, a wireless communication unit 14, a GPS (Global Positioning System) unit 16, an inertial sensor unit 18, an image processing unit 20, and drive control. It has a unit 22 and a power supply unit 24.

The mobile unit 1 can communicate with the base station 50 and the client device 60 by the wireless communication unit 14. The mobile unit 1 receives the information indicating the search target such as the data 70 from the base station 50 or the client device 60 by the wireless communication unit 14. The base station 50 is an example of a management device that manages the operation of the mobile unit 1, and appropriately gives an instruction regarding the operation. The client device 60 is, for example, a server arranged in a security room of a theme park, a server of a security company, a server of a police organization, or a mobile terminal of a person in charge of those organizations.

The data 70 illustrated in FIG. 3 is an example of search target information indicating search target information. The data 70 includes an image to be searched and information indicating its category. The category is a group to which a specific object belongs, such as “human”, “dog”, and “box”.

The GPS unit 16 and the inertial sensor unit 18 allow the mobile unit 1 to measure the position of the mobile unit 1 itself. The GPS unit 16 basically receives radio waves from three or more GPS satellites and measures the position of the mobile body 1. The inertial sensor unit 18 measures the position of the moving body 1 by integrating the movement of the moving body 1 from the starting point by using, for example, an acceleration sensor and a gyro sensor.

The image processing unit 20 allows the mobile unit 1 to operate the camera 3 and acquire an external image.

The moving body 1 controls the operation of each tire 7 by the drive control unit 22.

The power supply unit 24 is, for example, a replaceable rechargeable battery, and supplies power to each unit of the mobile body 1.

In the storage unit 12, in addition to various data and programs necessary for unmanned movement such as data indicating a movement plan for autonomously moving from a starting point to a destination position, information indicating the topography of a moving region and the position of a structure, The following programs are stored.

The storage unit 12 stores an information reception program, an image acquisition program, an object recognition program, a position information generation program, and a transmission information generation program.
The object recognition program includes a general object recognition program that recognizes a category of an object, and a specific object recognition program that determines whether a specific object is the same as a search target.

The CPU 10 and the information receiving program are examples of an information receiving unit that acquires search information from the outside. The mobile unit 1 receives the data 70 of FIG. 3 from the base station 50 of FIG. 2, for example. The data 70 is an example of search information.

The CPU 10 and the image acquisition program are an example of an image acquisition unit that operates the image processing unit 18 to acquire an image. The mobile unit 1 is adapted to continuously acquire images after receiving the search information.

The CPU 10 and the object recognition program are examples of object recognition means. The object recognition program includes feature data generated by deep learning. Deep learning is machine learning of a neural network having a multi-layer structure, and the field of image recognition is one of the most useful fields of application.

The moving body 1 can identify the feature of the object included in the image acquired by the camera 3 and recognize the category of the object based on a large number of feature information for each category. This category of recognition is called general object recognition.

The moving body 1 can also determine the identity of the object by determining the correlation between the image of the search target included in the search target information and the feature of the image of the object acquired by the camera 3 by the object recognition program. Has become. This determination of identity is called specific object recognition.

For example, as shown in FIG. 1, when an object such as a ferris wheel 100 and adults 200, 206, 208 and children 202, 204 are located in a theme park, the adults 200, children 202, etc. are treated as "humans". The recognition process is general object recognition. Then, the process of determining that the adult 200 and the person included in the data 70 have the same identity is the specific object recognition.

In general object recognition, with respect to an image acquired by the camera 3, for example, a large number of features such as contours and directions of individual configurations are extracted, and compared with the features of each category acquired by deep learning to determine the correlation (correlation degree). To judge. The higher the degree of correlation, the higher the possibility that the object in the acquired image is the same as the category to be searched. For example, when the correlation degree is 0, the possibility that the category is the same as the search target (hereinafter, referred to as “category common probability”) is 0, and when the correlation degree shows the maximum value, the category common probability is set. Is defined as 100%. The mobile body 1 determines that the category of the object in the acquired image is the same as the search target category when the category common probability is a predetermined reference value, for example, 95% or more.

In the specific object recognition, the moving body 1 determines that an object belonging to the same category as the search target (hereinafter, referred to as “inspection target”) has a high possibility of being the search target (hereinafter, referred to as “candidate object”). Call).

The mobile unit 1 determines the degree of correlation between the characteristics of the image of the data 70 that is the search target and the characteristics of the image of the inspection target. For example, with respect to the data 70, a large number of features such as the shape and direction of the broken eye, the centerline of the nose and the angle of the nose are extracted, and similarly, a large number of the features are also extracted for the inspection target. Judge that the degree of correlation is high. The higher the degree of correlation, the higher the possibility of being a search target. When the correlation degree is 0, the possibility of being a search target (hereinafter referred to as “probability”) is set to 0, and when the correlation degree shows the maximum value, the accuracy is defined as 100%.

The mobile body 1 determines that the inspection target is the search target when the accuracy is a predetermined determination reference value, for example, 90% or more, and sets it as a candidate object. The moving body 1 compares the characteristics of the data 70 with the characteristics of the image of the adult 200, and if the degree of correlation is high and the accuracy is 90% or more, it is regarded as a candidate object.

The CPU 10 and the position information generation program are examples of position information generation means.
The mobile unit 1 continuously generates and holds information indicating the absolute position and orientation of the mobile unit 1 by the GPS unit 16 and the inertial sensor unit 18. Further, a plurality of images of the same object are acquired by the camera 3, and the relative position with respect to the moving body 1 and the object such as the ferris wheel 100 (the position is known) of the adult 200 or the like which is a candidate object. The relative position of can be calculated. Since the absolute position of the mobile unit 1 itself can be measured by the GPS unit 16 and the inertial sensor unit 18, the mobile unit 1 refers to the absolute position of the mobile unit 1 itself to calculate the absolute position of the adult 200 or the like. You can A detailed description of the position measurement will be omitted.

The CPU 10 and the transmission information generation program are examples of transmission information generation program generation means. As shown in FIG. 4, the mobile unit 1 generates the transmission information 80 including the image of the adult 200 and the position information.

Since the storage unit 12 of the moving body 1 has the position information of the structures in the theme park and is known, the reference information indicating the position of the adult 200 in relation to the structure of the theme park is generated. You can The mobile unit 1 includes, for example, information "100m south of the Ferris wheel" in the transmission information 80. The transmission information 80 also includes information about the time when the moving body 1 acquires the image of the adult 200 and the probability that the adult 200 is a search target.

The mobile unit 1 transmits the transmission information 80 to the base station 50 or the client device 60 via the wireless communication unit 14 and the communication satellite 40.

The operation of the mobile unit 1 will be described below with reference to the flowchart of FIG. First, the mobile unit 1 determines whether or not the search target information has been received (step ST1). The search target information is, for example, the data 70 shown in FIG. In the case of the data 70, the category is “human”. When the mobile unit 1 determines that the search data 70 has been received, the mobile unit 1 extracts features from the data 70 (step ST2). Subsequently, the moving body 1 acquires an image of the object in the theme park of FIG. 1 while moving (step ST3), and determines whether or not the object in the image is the same as the search target (step ST3). ST4). In step ST4, general object recognition for recognizing whether or not the object in the image is the same as the search target category "human" and whether or not the object belonging to "human" is the search target is determined. Perform specific object recognition.

In step ST4, if the probability (accuracy) that the adult 200 is a search target is equal to or greater than a predetermined reference value, the moving object 1 regards the adult 200 as a candidate object and generates position information indicating the position of the candidate object (step S4). ST5). Subsequently, the mobile unit 1 generates transmission information 80 (see FIG. 4) including the image of the adult 200 and position information (step ST6), and transmits the transmission information 80 to the base station 50 or the client device 60 (step ST7).

In the above, an example in which the position of the candidate object is included in the transmission information 80 has been described, but as a simple method, the position of the moving body 1 may be used as position information indicating the approximate position of the candidate object. Alternatively, the position information of the moving body 1 may be included in the transmission information and transmitted, and the position of the candidate object may be transmitted later. Further, the search target is not limited, and for example, in addition to the above-mentioned criminal suspect, a child lost in the theme park, a runaway person, or a disappeared person may be used.

<Second embodiment>
In the second embodiment, a moving vector calculation program is stored in the storage unit 12 of the moving body 1 to calculate the moving direction and moving speed of the candidate object. In addition, a transmission information size adjustment program is stored in the storage unit 12 of the mobile unit 1 and adjusts the data size of the transmission information. The CPU 10 and the movement vector calculation program are an example of movement vector calculation means. The CPU 10 and the transmission information size adjusting program are examples of the transmission information size adjusting means.

The object may be moving at time t1 when the image of the candidate object is acquired. In the second embodiment, such a moving state is also included in the transmission information. The movement vector can be calculated by referring to the image at the time t1 and the image acquired at the time t1+α, which is a little after the time t1. Note that the movement of the object can be easily measured if the moving body 1 is stopped and the position is fixed during the movement vector measurement. Alternatively, when the moving body 1 is moving, the true movement of the object can be measured by offsetting the movement of the moving body 1 from the apparent movement of the object.

For example, as shown in FIG. 6, the moving body 1 includes, in the transmission information 82, movement information that the adult 200, which is a candidate object, is heading south-southwest at a speed of 10 km (km) per hour.

The mobile unit 1 adjusts the data size of the transmission information according to the type of device that receives the transmission information and the communication environment including the state of the radio wave at that time. For example, when transmitting the transmission information to the client device 60 (FIG. 2), if the client device 60 is a mobile terminal, the data size is reduced, and as shown in the transmission information 82 of FIG. Only the image data is included in the transmission information. Alternatively, when the radio wave is weak or the transmission and reception of the radio wave is unstable, the image data of only the face portion is included in the transmission information. On the other hand, when the client device 60 is a server or when the communication environment is good, as shown in FIG. 4, the image data of the whole body is included in the transmission information. Alternatively, if the transmission information includes image data of only the face portion of the candidate object and a request is received from the base station 50 or the client device 60, the image data of the whole body is transmitted in response to the request. You may make it the structure.

<Third embodiment>
In the third embodiment, the storage unit 12 of the moving body 1 stores a position prediction program in addition to the movement vector calculation program of the second embodiment, and stores the position of the candidate object after a predetermined time. calculate. The CPU 10 and the position prediction program are examples of position prediction means.

For example, as shown in the transmission information 84 in FIG. 7, the mobile unit 1 transmits the latitude and longitude after 15 minutes as the expected position after 15 minutes, and the information “South Gate” in the theme park as reference information. Include in information.

<Fourth Embodiment>
For example, the appearance of a criminal suspect may change during flight. Assuming that the person to be searched is the person shown in the image 70 of FIG. 3, when the moving body 1 searches in the theme park, for example, wearing glasses (FIG. 8(a)), wearing eyelashes, lipstick, etc. May be present (FIG. 8(b)). In addition, the hairstyle may be changed (FIG. 8C) to disguise. Alternatively, the face may be damaged due to great effort during the escape (FIG. 8D). In this way, when the appearance of the search target changes, the search target exists in the theme park, and even if an image of the search target is acquired, the degree of correlation regarding the identity decreases.

Therefore, the mobile unit 1 according to the third embodiment has a search target deep learning program for acquiring data showing characteristics of the search target itself by deep learning. If a large number of image data other than the data 70 (FIG. 3) is acquired for the search target, the mobile unit 1 performs machine learning using the image data. If the image data 70 is the only image data, the image data 70 may be processed into, for example, the images shown in FIGS. 8A to 8D, or the outline may be blurred. , Machine learning is performed after generating a plurality of image data. As a result, since the search target itself can be generalized, changes in the appearance of the search target can be easily accommodated.

The operation of the mobile unit 1 will be described below with reference to the flowchart of FIG. As shown in FIG. 9, when the mobile body 1 acquires the search target information (step ST1), the mobile body 1 performs deep learning on the search target and then extracts a feature (step ST2). Since step ST3 and subsequent steps are the same as those in the first embodiment, description thereof will be omitted.

Note that the search target deep learning program may be held by the base station 50, and the base station 50 may perform deep learning on the search target. In this case, the mobile unit 1 receives and uses the result of deep learning on the search target performed by the base station 50.

<Fourth Embodiment>
The search target may not be a specific person or object, but may be a “human satisfying a specific condition” or an “object satisfying a specific condition”. "Human satisfying a specific condition" is a subcategory under the "human" category. The “person who meets a specific condition” is, for example, a person who is in poor physical condition or a suspicious person. The “object satisfying a specific condition” is, for example, a suspicious object. In this case, the search target information does not necessarily include an image, but may be text data indicating a subcategory “person with poor physical condition”.

That is, in the fourth embodiment, the moving body 1 performs general object recognition in the subcategory.

The object recognition program of the moving body 1 has characteristic data of “human satisfying a specific condition” and “object satisfying a specific condition” generated by deep learning.

For example, with regard to "person with poor physical condition", use a large number of images of people who are lying on the ground or a large number of people who are lying down and sitting on a chair, and classify them into lower subcategories as appropriate. , Do deep learning. The mobile unit 1 holds data indicating the characteristics acquired by deep learning. Then, when the mobile body 1 receives the text data “person with poor physical condition” as the search target information, the mobile body 1 acquires and refers to the data indicating the characteristics stored in the storage unit 12, and, for example, the theme park. It is determined that the person 208 is a “person with a poor target”. The human 208 is a fallen human.

As for “suspicious person”, for example, using a large number of images of people with unnaturally large luggage or long objects, people with hidden faces, or people with unnaturally thick clothes, Deep learning is performed by classifying into sub-categories. The mobile unit 1 holds data indicating the characteristics acquired by deep learning. The sub-categories below are, for example, "people wearing unnatural clothes", "people carrying unnaturally large luggage", "people wearing gloves even in the summer", and "sunglasses despite being cloudy". The person who wears it." When the mobile body 1 receives the text data “suspicious person” as the search target information, the mobile body 1 acquires and refers to the data indicating the characteristics of the “suspicious person”. It is judged that it is. The human 206 is unnaturally thick and has an explosive material 206a hidden inside the clothes.

For suspicious objects, deep learning is performed by using a large number of images of exposed bombs and firearms, packed bombs and firearms, and appropriately classifying them into subordinate categories. The mobile unit 1 holds data indicating the characteristics acquired by deep learning. When the mobile body 1 receives the text data “suspicious object” as the search target information, the mobile body 1 acquires the data indicating the characteristics of the “suspicious object” and refers to the text, for example, a box in the theme park. It is determined that 210 is a “suspicious item”. The box 210 is unnaturally sealed with gum tape, and the explosive 210a is stored inside.

Hereinafter, the operation of the mobile unit 1 will be described with reference to the flowchart of FIG. Mobile unit 1
When the search target information is acquired (step ST1), the search target feature is acquired (step ST2). The search target is, for example, a subcategory of “person with poor physical condition”. The information indicating the characteristics of the search target is generated in advance by deep learning and stored in the storage unit 12. Then, the mobile body 1 acquires the image in the theme park (step ST3), and judges whether it corresponds to a "person with poor physical condition" (step ST4). Since step ST5 and subsequent steps are the same as those in the first embodiment, description thereof will be omitted.

In step ST4, for example, when an image of a person who is highly likely to be a search target is acquired a plurality of times at intervals of a predetermined time, and the condition “person with poor physical condition” is met for a predetermined time Alternatively, it may be determined that the person is “a person who is in poor physical condition”.

When the search target is a “suspicious person” or a “suspicious object”, the reference value serving as a reference for the candidate object may be set low. For example, a suspicious person may behave invisible to the suspicious person, and may have the appearance of not appearing suspicious when explosives are left behind in a theme park. In this case, the degree of correlation between the image acquired by the camera 3 and the search target is low, but even in such a case, it is desirable to transmit the transmission information to the base station 50 or the client device 60.

<Fifth Embodiment>
In the fifth embodiment, rather than a relatively narrow space such as a theme park,
The mobile body searches a wide area. In the fifth embodiment, as shown in FIG. 11, the moving body is an unmanned aerial vehicle 300, a so-called multicopter or drone, and has a plurality of rotary wings. In the example of FIG. 11, the unmanned aerial vehicle 300 has a main body 302 and four rotors 304. Each rotor 304 is connected to a motor. In the main body 302, a communication device, an autonomous flight control device, a GPS device, an inertial sensor such as an acceleration sensor and a gyro sensor, a barometer, a battery and the like are arranged. A camera 306 is connected to the main body 302. The camera 306 is a digital camera that captures a digital image, a video camera, or the like. The unmanned aerial vehicle 300 can communicate with the base station 50 (see FIG. 2) and the client device 60 directly or through the communication satellite 40 by a communication device. The functional blocks of the unmanned aerial vehicle 300 are as shown in FIG. 2, similarly to the mobile body 1 of the first embodiment.

In FIG. 11, humans 250 and 252 are victims in the mountainous area 310. The human 254 is a searcher (hereinafter, referred to as “searcher 254”) searching for a victim. The searcher 254 wears a mark 254a indicating that he is a searcher on his hat and sleeve, so that the unmanned aerial vehicle 300 can distinguish the victim and the searcher. When the unmanned aerial vehicle 300 receives the information of the subcategory “distressed person” as the search target information, it determines that a person other than the searcher 254 is a “distressed person” in the designated area. It has become. That is, the image including the mark 254a is excluded, and the “human” not including the mark 254a is determined to be the victim.

The operation of the unmanned aerial vehicle 300 will be described below with reference to the flowchart of FIG. When the unmanned aerial vehicle 300 acquires the search target information (step ST1), the unmanned air vehicle 300 acquires the characteristics of the search target information (step ST2). The search target information is, for example, “distressed person”. In the case of the “distressed person”, the characteristic of the search target information is a person located in a predetermined area, other than the searcher 254. Subsequently, the unmanned aerial vehicle 300 acquires an image (step ST3), and if the category is “human”, it is determined whether or not the person is the searcher 254a, that is, whether or not it is an exclusion target (step ST4), and the exclusion target. If not, the process proceeds to step ST5. Since step ST5 and subsequent steps are the same as those in the first embodiment, description thereof will be omitted.

Although the unmanned air vehicle 300 has been described above, for example, when searching for a victim on the sea or in a lake, an unmanned ship or unmanned boat may have the same function as the unmanned air vehicle 300.

In the first to fifth embodiments described above, programs such as an image acquisition program, an object recognition program, a position information generation program, and a transmission information generation program are stored in the storage unit 12 of the mobile body 1 or the unmanned air vehicle 300. However, it is configured such that some or all of the above programs are stored in the storage unit in the server of the base station 50 and the results processed by the base station 50 are used by the mobile unit 1 and the unmanned aerial vehicle 300. You may. In this case, the storage unit 12 of the mobile body 1 or the unmanned aerial vehicle 300 only needs to store a processing program for processing an instruction from the base station 50. Further, some or all of the characteristic data may be stored in the storage unit in the server of the base station 50, and the mobile unit 1 may be configured to acquire the characteristic data as needed.

The monitoring device and the monitoring method of the present invention are not limited to the above-described embodiment, and various changes can be made without departing from the scope of the present invention.

1 Mobile 10 CPU
14 wireless communication unit 16 GPS unit 18 inertial sensor unit 20 image processing unit 22 drive control unit 24 power supply unit 300 unmanned air vehicle

Claims (8)

A monitoring device configured as a mobile body,
An information receiving means for receiving search target information indicating a search target,
An image acquiring means for acquiring an image of the object body,
An object recognition unit that determines whether the object included in the image is the search target,
Position information generating means for generating position information indicating a position of the object which is a position different from the position of the moving body ,
Transmission information generating means for generating transmission information including the image of the object and the position information,
And a monitoring device. The monitoring device according to claim 1, further comprising a transmission information size adjusting unit that adjusts a data size of the transmission information according to a type of a device that receives the transmission information or a communication environment including a radio wave state. The object recognition unit, viewed contains the data generated by the deep learning (deep learning),
General object recognition for determining whether the object is an inspection target that is the object belonging to the same category as the search target, and specific object recognition for determining whether the inspection target is the search target. Do,
The monitoring device according to claim 1 or 2. The position information generating means, for an object that is likely to be the search target, a relative position to the absolute position of the moving body that is continuously calculated, and a known position of a structure in a region where the moving body moves. Position information indicating the position of the object based on the relative position of
The monitoring device according to any one of claims 1 to 3 . The search target is a suspicious person or a suspicious object,
The object recognition means sets a determination reference value for determining that the object is the search target, lower than a determination reference value when the search target is other than the suspicious person or the suspicious object,
The monitoring device according to any one of claims 1 to 4. A monitoring device configured as a mobile
An information receiving step of receiving search target information indicating a search target,
An image acquisition step of acquiring an image of the object,
An object recognition step of determining whether the object included in the image is the search target;
For the object that is likely to be the search target, based on the relative position of the absolute position of the moving body continuously calculated and the known position of the structure of the region where the moving body moves, A position information generating step of generating position information indicating the absolute position of the object ,
A transmission information generating step of generating transmission information including the image of the object and the position information,
A monitoring method having. In the object recognition step,
General object recognition for determining whether the object is an inspection target that is the object belonging to the same category as the search target, and specific object recognition for determining whether the inspection target is the search target. Do,
The monitoring method according to claim 6 . Prior to the transmission information generation step,
A transmission information size adjusting step of adjusting the data size of the transmission information according to the type of device that receives the transmission information, or the communication environment including the radio wave state,
The monitoring method according to claim 6 or 7 .