KR20240068826A - Ai-harmful animal monitoring apparatus using drone and eradication attack method using the same - Google Patents

Abstract

Our center uses artificial intelligence to detect harmful animals approaching the farm and exterminate them. More specifically, it is equipped with a drone thermal imaging camera and expulsion attack means on the drone, and uses an artificial intelligence neural network on a mobile device. Through this, harmful animals exposed to farm images and drone images are detected, and when the locations of the animals are specified on the farm grid map, the drone is automatically dispatched to the corresponding coordinates and transmits the drone image of the location to the farm owner's mobile device. It is about an AI-pesticide surveillance device using a drone, which retransmits the drone and activates the drone's extermination attack method according to a remote command from the farm owner, and an extermination method using the same.

Description

AI-harmful animal monitoring device using drones and extermination attack method using the same {AI-HARMFUL ANIMAL MONITORING APPARATUS USING DRONE AND ERADICATION ATTACK METHOD USING THE SAME}

This center uses artificial intelligence to detect harmful animals approaching the farm and exterminate them. More specifically, it is equipped with a drone thermal imaging camera and expulsion attack means on the drone, and an artificial intelligence neural network on a mobile device. Through this, harmful animals exposed to farm images and drone images are detected, and when the locations of the animals are specified on the farm grid map, the drone is automatically dispatched to the corresponding coordinates and transmits the drone image of the location to the farm owner's mobile device. It is about an AI-pesticide surveillance device using a drone, which retransmits the drone and activates the drone's extermination attack method according to a remote command from the farm owner, and an extermination method using the same.

With the recent development of artificial intelligence technology, artificial intelligence technology is applied to harmful animal extermination technology and is expected to develop rapidly.

One of the known harmful animal extermination technologies is to set in advance a unique ultrasonic frequency band that responds sensitively to each animal to be exterminated and to exterminate it using ultrasonic waves when the animal is detected using a PIR sensor (Passive Infrared Sensor). Since it uses ultrasonic waves, it has the advantage of being able to exterminate harmful animals without harming humans. However, since it does not operate by distinguishing types of harmful animals, not only is it impossible to respond according to the type of animal, but also the movement of insignificant animals is affected. There is also a disadvantage that the ultrasonic device malfunctions sensitively, greatly reducing its efficiency, and over time, the animal's immune response to ultrasonic waves increases, drastically reducing the repelling effect.

Another known representative harmful animal extermination technology is to receive images of the area using a camera, detect wild animals using a pre-trained artificial intelligence neural network, and step by step use lights, speakers, and repellent to protect them from wild animals. As a technology to exterminate animals, it has the advantage of being able to exterminate animals by distinguishing between types of animals.

However, these techniques make it difficult to clearly distinguish between animals when the weather is cloudy or dark at night, and if the animal moves into the camera's blind spot during the extermination process, further tracking is impossible, and the animal is not afraid of lights, speakers, or repellent. In the case of self-adaptation, there is a problem that it is difficult to expect any further eradication effect using existing eradication methods.

In addition, because the range in which the extermination device operates is limited, as the surveillance area expands, the number of extermination devices increases rapidly, and maintenance and repair costs increase, which has the disadvantage of drastically reducing cost-effectiveness.

As damage caused by harmful animals on farms is spreading globally and its importance is increasing day by day, there is an urgent need for devices with accurate animal recognition capabilities and continuous and effective harmful animal extermination attack methods.

Republic of Korea Patent No. 10-1937597, which is the background technology of this application, relates to a drone for exterminating harmful animals and its flight method. The above patent discloses a drone for repelling harmful animals, which is equipped with sound, lighting, and a striking launcher as a means of repelling.

This institute aims to solve the problems of the prior art described above, and collects images through thermal imaging cameras installed on farms and drones to monitor the presence of harmful animals not only during the day but also at night, and receives farm or drone images. When a harmful animal is detected by an animal recognition artificial intelligence neural network on a mobile device that can analyze and recognize the type of harmful animal and the animal's eyes and body, the drone automatically dispatches to the location and tracks the harmful animal at the same time. Provides an AI-pesticide monitoring device using drones that exterminates by extermination attack means.

However, the technical challenges sought to be achieved by the embodiments herein are not limited to the technical challenges described above, and other technical challenges may exist.

According to one implementation of the present application, a thermal imaging camera can collect both thermal imaging images and general images, and transmits the images in real time to the farm manager's mobile device so that the farm manager can check them, but is not limited to this. no.

According to one embodiment of the present application, since a thermal imaging camera detects living things using infrared wavelength light that occurs, harmful animals can be easily detected day and night by using a thermal imaging camera that provides both general images and thermal images. It can be recognized.

According to one embodiment of the present application, a farm thermal imaging camera is placed in a predetermined area where harmful animals are to be exterminated to monitor the presence of harmful animals, and uses a farm grid map that can specify the location of the farm to detect harmful animals. One or more cameras may be installed in different locations to determine the indicated location, but are not limited to this.

According to one embodiment of the present application, the presence of harmful animals appearing on the farm is monitored by a drone that regularly patrols without a farm thermal imaging camera, and the location of the harmful animals appears using a farm grid map that can specify the location of the farm. It may be possible to determine, but is not limited to this.

According to one implementation of the present application, the farm manager's mobile device may include a mobile application for remotely checking the image of the thermal imaging camera in real time on the mobile device or activating a drone extermination attack method.

According to an implementation example of the present application, the animal recognition artificial intelligence neural network receives farm images or drone images generated from a thermal imaging camera as input, and creates a bounding box for the area where harmful animals are detected through a pre-trained artificial intelligence neural network. It may include an animal area detection unit that provides a bounding box, and an animal semantic segmentation unit that provides labeled images for each pixel of the harmful animals through a pre-trained artificial intelligence neural network.

It is preferred that the animal semantic segmentation unit classifies animals differently depending on the type of animal and the eyes and body of the animal, and performs labeling on a pixel basis.

According to an embodiment of the present application, an animal recognition artificial intelligence neural network, an animal determination unit, an animal position coordinate calculation unit, and an eye position coordinate determination unit can be installed on a drone or mobile application, and the drone or mobile application can be installed on a drone or mobile application from a drone thermal imaging camera. By analyzing the video, you can identify the animal type, location, eyes and body of the animal, and activate appropriate expulsion attack methods to attack harmful animals.

According to one implementation of the present application, the remote server includes an animal recognition artificial intelligence neural network, an animal determination unit, an animal position coordinate calculation unit, and an eye position coordinate determination unit, and the remote server remotely receives a drone image to determine the type, location, and information about the animal. By identifying the eyes and body of the animal and transmitting the corresponding drone control commands to the drone wirelessly, the drone's means of extermination and attack can be controlled.

According to one embodiment of the present application, the animal area detection unit may use YOLO (You Only Look Once), an artificial intelligence neural network, but is not limited thereto.

According to one embodiment of the present application, the drone's extermination attack means may include, but is not limited to, a threatening flight accompanied by sound or ultrasonic waves and a laser attack aimed at the animal's eyes.

According to one embodiment of the present application, a drone may be equipped with an ultrasonic speaker for attacking the animal with an ultrasonic frequency band appropriate for the type of harmful animal while flying in pursuit of the harmful animal.

According to an embodiment of the present application, the farm manager checks the drone image with a mobile device, selects one or more attack methods selected from among the extermination attack methods consisting of laser, ultrasonic waves, sound, and threat flight, and remotely instructs the drone to attack. A mobile application can be provided.

The sounds may include, but are not limited to, the sound of lightning, the sound of a hunting dog barking, the roar of a tiger, the sound of a gun, or the sound of a person speaking.

According to one implementation of the present application, all images captured by the farm thermal imaging camera and the drone thermal imaging camera can be received by the farm manager's mobile device, and the farm manager watches these images and sends a remote control command to attack the drone. The means can be remotely controlled and used.

According to one implementation of the present application, when the farm manager activates the extermination attack means using a mobile application, the drone kills harmful animals using the pre-activated expulsion attack means on the drone without a remote control command from the farm manager. You can get rid of it yourself.

According to one implementation of the present application, images transmitted to a mobile device can be stored in memory for a certain period of time, playback speed can be adjusted or rewinded, and the name, date, and time of the camera are displayed on each image. Controlled storage is preferred.

According to an implementation example of the present application, the control unit of the mobile application includes a drone control command transmitter that transmits a series of drone control commands to exterminate animals to the drone when the farm thermal imaging camera and the drone thermal imaging camera detect harmful animals. can do.

The above-described means of solving the problem are merely illustrative and should not be construed as intended to limit the present application. In addition to the exemplary embodiments described above, additional embodiments may be present in the drawings and detailed description of the invention.

According to the above-described means of solving the problem of this institute, the AI-harmful animal detection device using a drone according to this institute is capable of monitoring harmful animals at all times, day and night, using thermal imaging images through an animal recognition artificial intelligence neural network, and is equipped with expulsion attack means. Tracking and extermination of harmful animals can be carried out automatically using a single drone, providing farm owners with the convenience of monitoring and eradicating harmful animals.

In addition, the animal recognition artificial intelligence neural network can be installed on a drone or mobile device, and in this case, it can recognize the type of harmful animal and the animal's eyes and body while tracking harmful animals in real time by analyzing drone images. , During the tracking process, the most vulnerable part of the harmful animal's body, the eye, can be targeted and directly attacked with a laser, thereby maximizing the extermination effect.

In addition, extermination attacks include not only lasers but also low-level attacks such as lightning sounds, tiger roars, gunshots, and human speaking sounds, which has the advantage of being able to exterminate harmful animals without causing serious harm.

In addition, since the ultrasonic frequency band appropriate for the type of harmful animal classified by the animal recognition artificial intelligence neural network is automatically selected to attack the animal, it has the advantage of reducing the immune response of harmful animals to ultrasonic waves.

In addition, expulsion attack means can be mounted on the drone to chase harmful animals, perform threatening flights accompanied by sound or ultrasonic waves, and laser attacks aimed at the animal's eyes. These expulsion attack means can be deployed in various stages by stage. Eradication operation can be carried out

In addition, all images captured by farm thermal imaging cameras and drone thermal imaging cameras can be received by the farm manager's mobile device, and the farm manager can activate or remotely control the drone's extermination attack means through remote control commands. .

However, the effects that can be obtained herein are not limited to the effects described above, and other effects may exist.

Figure 1 is a conceptual diagram of an AI-harmful animal monitoring device using a drone according to an implementation example of the present application.
Figure 2 is a conceptual diagram of a mobile application according to an implementation example of the present application.
Figure 3a is an embodiment showing an example of the operation of an animal recognition artificial intelligence neural network composed of an animal determination unit and an eye position coordinate determination unit according to the present application.
Figure 3b is an example of an animal's eye position coordinate determination unit according to the present application.
Figure 3c shows several embodiments of a drone-based repelling attack means according to the present application.

Below, with reference to the attached drawings, embodiments of the present application will be described in detail so that those skilled in the art can easily implement them. However, the present application may be implemented in various different forms and is not limited to the embodiments described herein. In order to clearly explain the present application in the drawings, parts that are not related to the description are omitted, and similar parts are given similar reference numerals throughout the specification.

Throughout this specification, when a part is said to be “connected” to another part, this includes not only the case where it is “directly connected,” but also the case where it is “electrically connected” with another element in between. do.

Throughout this specification, when a member is said to be located “on”, “above”, “at the top”, “below”, “at the bottom”, or “at the bottom” of another member, this means that a member is located on another member. This includes not only cases where they are in contact, but also cases where another member exists between two members.

Throughout the specification of the present application, when a part "includes" a certain element, this means that it may further include other elements rather than excluding other elements, unless specifically stated to the contrary.

As used herein, the terms "about", "substantially", etc. are used to mean at or close to the numerical value when manufacturing and material tolerances inherent in the stated meaning are presented, and to aid understanding of the present application. It is used to prevent unscrupulous infringers from unfairly exploiting disclosures in which precise or absolute figures are mentioned. Additionally, throughout the specification herein, “a step of” or “a step of” does not mean “a step for.”

Throughout this specification, the term "combination thereof" included in the Markushi format expression means a mixture or combination of one or more components selected from the group consisting of the components described in the Markushi format expression, It means including one or more selected from the group consisting of.

Throughout this specification, description of “A and/or B” means “A or B, or A and B.”

Throughout the specification of this application, harmful animals may include not only wild animals but also all living and moving animals, such as humans and birds, that exist within a predetermined designated area.

Throughout this specification, thermal imaging cameras collectively refer to farm thermal imaging cameras and drone thermal imaging cameras.

The wireless transmission/reception connection of the present invention is preferably made by Wifi or Internet of Things (IoT) connection.

Throughout the specification herein, a farm grid map refers to a map that displays location information within a farm in the form of coordinates that can be generated by one or more farm cameras or reconnaissance drones.

Throughout this specification, farm owner is used interchangeably with farm manager.

Throughout the specification herein, farm images refer to images collected from a plurality of farm thermal imaging cameras installed within the farm.

Throughout this specification, drone image refers to images collected from a drone thermal imaging camera mounted on a drone.

Hereinafter, the AI-harmful animal monitoring device using a drone and the extermination attack method using the same will be described in detail with reference to implementation examples, examples, and drawings. However, the present application is not limited to these embodiments, examples, and drawings.

Figure 1 is a conceptual diagram of an AI-harmful animal monitoring device using a drone according to an implementation example of the present application.

As a technical means for achieving the above-mentioned technical problem, the first aspect of the present application is an AI-pest surveillance device using a drone including a mobile application 70 and a drone 30 installed on a mobile device 40. , the drone 30 includes a drone thermal imaging camera 31 for photographing animals during aerial flight and a repelling attack means 32 for repelling and attacking animals, and the mobile application 70 captures the drone thermal imaging. An animal recognition artificial intelligence neural network 10 for detecting the eyes and body of an animal exposed to the camera 31, and an attack means selection menu that can remotely select and activate the expulsion attack means through a touch screen on the mobile device. Provides an AI-harmful animal monitoring device using a drone, including (70c).

The repelling attack means 32 of the drone 30 includes a laser generator 32a, an ultrasonic speaker 32b, and an acoustic speaker 32c, and the repelling attack method using the repelling attack means 32 includes an ultrasonic speaker ( Threat flight accompanied by the ultrasonic waves of 32b) or the sound of the acoustic speaker 32c; and a laser attack aimed at the animal's eyes using the laser generator 32a, and the repellent attack means can be activated according to a remote command from the farm owner on the mobile application 70.

The mobile application 70 includes an attack means selection menu 70c that can be activated by remotely selecting an expulsion attack method through a touch screen on the mobile device 40 and a sound selection menu 70d for selecting a sound type. It can be included.

The attack method selection menu 70c preferably includes buttons for selecting and activating attack methods including threat flight, ultrasound, sound, and laser.

The sound selection menu 70d preferably includes buttons for selecting one or more sounds among a tiger roar, people talking, a dog barking, a car horn, and a gun sound.

In addition, the mobile application 70 allows the farm manager to check the drone video screen 70a and the laser pointing screen 70b in real time through the mobile application 70 and press the laser firing button 70e on the mobile application 70. You can press it. It is preferred that each time the laser firing button 70e is pressed, one laser is fired from the laser generator 32a. In this case, the accuracy of the laser attack can be increased by pressing the laser button after confirming that the laser is accurately aimed at the animal's eyes using the laser pointing screen 70b.

The drone image screen 70a is a screen that shows real-time images captured by the drone thermal imaging camera 31 mounted on the drone 30 and is operated in conjunction with the mobile application 70.

The laser pointing screen 70b is a screen that displays the current location of the laser beam over the real-time image captured by the drone thermal imaging camera 31 mounted on the drone 30, and is linked with the mobile application 70. It runs.

In addition, the mobile application 70 has a remote attack mode for the farm manager to remotely control the flight of the drone 30 or remotely control the extermination attack means 32, or the drone 30 can use the expulsion attack means on the drone itself. It may include a remote attack mode button (70f) and an automatic attack mode button (70g) for selecting whether to use the automatic attack mode to exterminate harmful animals.

In the case of the automatic attack mode, it is preferred that the drone 30 uses sound and the repellent attack means 32 activated in advance by the farm manager through the mobile application 70 to exterminate harmful animals.

On the other hand, in the case of remote attack mode, the drone 30 executes the extermination attack means 32 or sound activated in advance through the mobile application 70 once every time the farm manager presses the remote attack mode button to kill harmful animals. It is preferred to exterminate, in which case the automatic attack mode is preferably canceled automatically.

In addition, the farm image obtained by the farm thermal imaging camera 2 can be provided to the farm manager's mobile device 40 by the wireless transmitter 3, and the farm manager can view the farm image in real time through the mobile application 70. You can check the drone video.

The communication connection between the drone 30 and the mobile device 40 is preferably made via Wi-Fi or wireless Internet.

When the farm thermal imaging camera 2 and the drone thermal imaging camera 31 detect harmful animals, the mobile application 70 issues a drone control command so that the drone 30 is dispatched to the location coordinates of the animals and performs an extermination attack. It is transmitted to the drone (30).

The drone control command may include selecting the expulsion attack means 32, activating the expulsion attack means 32, a drone dispatch command, and a real-time remote attack command. For example, a farm manager can attack harmful animals by remotely controlling the extermination attack means 32 on the drone 30 through a real-time remote attack command while watching the real-time drone image 70a or the laser pointing screen 70b.

In addition, the mobile application 70 can check farm images and drone images in real time, and includes animal location coordinates, animal type, eye location coordinates, number of animals, animal appearance time, and eye location coordinate information, etc. can be calculated.

Figure 2 is a conceptual diagram of a mobile application 70 according to an implementation example of the present application.

Referring to FIG. 2, the mobile application 70 includes an animal recognition artificial intelligence neural network 10 for detecting the eyes and body of an animal exposed to a farm thermal imaging camera 2 or a drone thermal imaging camera 31; an animal location coordinate calculation unit 54 for specifying the location coordinates of animals detected on the farm grid map 52; and a control unit 20.

In addition, the control unit 20 includes an animal determination unit 22 that determines the type of animal and the number of animals from the results of the animal recognition artificial intelligence neural network 10; an eye position coordinate determination unit (13) that determines the eye position of the animal from the results of the animal recognition artificial intelligence neural network (10); And drone control that wirelessly transmits a drone control command to the drone 30 to dispatch the drone 30 to the location coordinates of the animals, remotely control the drone 30 to perform an attack, or activate the expulsion attack means 32. It includes a command transmission unit 23.

The mobile application 70 can check farm images and drone images in real time, and determines the location coordinates of the animal by the animal recognition artificial intelligence neural network 10, the animal location coordinate calculation unit 54, and the control unit 20. The type, eye position coordinates, and number of animals can be calculated, or a drone control command can be wirelessly transmitted to the drone 30.

In the remote attack mode, the mobile application 70 selects one or more attack means selected from among the expulsion attack means 32 consisting of laser, ultrasonic waves, sound, and threat flight according to the selection of the farm manager and remotely attacks the drone 30. Attack orders can be given.

On the other hand, in the automatic attack mode, the drone 30 can exterminate harmful animals on its own according to the drone control command provided by the mobile application 70 without a separate remote control command from the farm manager.

That is, in the automatic attack mode, the drone 30 can independently exterminate harmful animals according to the drone control command using the expulsion attack means 32 that the farm manager activated in advance using the mobile application 70. In this case, it is preferred that the drone control command transmitter 23 selects an attack method according to the type or extermination scenario of the identified animal, sets up a drone control command, and wirelessly transmits it to the drone.

The animal recognition neural network 10 can be trained to detect the appearance of harmful animals through analysis of farm images from the farm thermal imaging camera 2.

Additionally, the animal recognition neural network 10 can be trained to detect the appearance of harmful animals through analysis of drone images from the drone thermal imaging camera 31.

The animal recognition artificial intelligence neural network 10 of the present invention is an artificial intelligence neural network that has been pre-trained by various images including harmful animals in order to detect the appearance of harmful animals through analysis of drone images or farm images, and has an animal area detection unit. (11) and an animal semantic segmentation unit (12).

The animal area detection unit 11 preferably uses the YOLO neural network model that provides a bounding box for the animal area to be found by an object detection method.

Through the animal semantic segmentation unit 12, it is possible to obtain images labeled by color by distinguishing the type of animal and the animal's body and eye area shown in the drone image or farm image.

The animal location coordinate calculation unit 54 calculates the positions of the animals on the farm grid map 52 using drone images or farm images, and uses the image of the regional farm grid map 52 and the drone on the coordinates where the animals appear. It is preferred to calculate the location coordinates of the animal by comparing the fitting similarity between images and farm images.

In the present invention, fitting similarity determination is performed using any one of the SSD (Sum of Squared Difference), SAD (Sum of Absolute Difference), KNN (K-nearest neighbor algorithm), and NCC (Normalized Cross Correlation) techniques that are performed on two-dimensional images. It is preferred to calculate using .

The animal recognition artificial intelligence neural network 10 includes an animal area detection unit 11 that provides a bounding box for the animal's body or eye area; and an animal semantic segmentation unit 12 that distinguishes the body and eyes of the animal and performs pixel-by-pixel labeling, and the control unit 20 is bounded by the animal area detection unit 11 and the animal semantic segmentation unit 12. An animal determination unit (22) that determines whether an animal exists within a bounding box and the type of animal; an eye position coordinate determination unit 13 that calculates the eye position of the animal using the animal area detection unit 11 and the animal semantic segmentation unit 12; and a drone control command transmission unit 23 that transmits a drone control command to the drone 30 so that the drone 30 is dispatched to the location coordinates of the animals provided by the animal location coordinate calculation unit 54 and performs an extermination attack. The drone control command includes the appearance location of the animal, animal type, appearance time and population information, eye position coordinate information, selection of the expulsion attack means 32, activation of the expulsion attack means 32, and command to dispatch the drone 30. , and may include real-time remote attack commands.

The animal determination unit 22 determines that the number of mark pixels of the animal body generated by the animal semantic segmentation unit 12 within the bounding box of the animal body generated by the animal area detection unit 11 is greater than a predetermined number. In this case, it is preferred to determine that the animal exists within the bounding box.

The animal determination unit 22 of the present invention combines the animal area detection unit 11 and the animal semantic segmentation unit 12 to determine whether an animal appears and the type of animal, thereby ensuring high accuracy.

In addition, another aspect of the AI-harmful animal monitoring device of the present invention includes an animal recognition artificial intelligence neural network (10), an animal determination unit (22), an animal position coordinate calculation unit (54), and an eye position coordinate determination unit (13). It is provided inside the drone 30, and the drone 30 itself analyzes the drone image to recognize the type of animal and the eyes and body of the animal, thereby automatically activating the repellent attack means 32. In this case, the drone 30 can track and exterminate harmful animals alone without the help of the mobile application 70.

Additionally, in this case, the mobile application 70 can receive the drone image from the drone 30 and provide it to the farm manager, and the farm manager can use the buttons on the mobile application 70 to control the drone ( 30) Harmful animals can be repelled by controlling the repelling attack means 32 on the top.

In addition, another aspect of the AI-harmful animal monitoring device of the present invention includes an animal recognition artificial intelligence neural network (10), an animal determination unit (22), an animal position coordinate calculation unit (54), and an eye position coordinate determination unit (13). It includes a remote server (not shown), and the remote server remotely receives and analyzes drone images captured by a drone, determines the type, location, and eyes and body of the animal through this, and gives a drone control command accordingly. It is possible to exterminate harmful animals by wirelessly transmitting to the drone 30 and controlling the expulsion attack means 32 on the drone 30.

Also, in this case, the mobile application 70 can receive the drone image from the remote server and provide it to the farm manager, and the farm manager can use the buttons on the mobile application 70 to send a remote control command to the drone for extermination attack means. You can exterminate harmful animals by controlling (32).

Figure 3a is an embodiment showing an example of the operation of an animal recognition artificial intelligence neural network composed of an animal determination unit and an eye position coordinate determination unit according to the present application.

Referring to FIG. 3A, the animal area detection unit 11 is a body area detection unit 11a that generates a bounding box for the animal body area in the drone image or farm image, and the position of the animal's eyes. It consists of an eye area detection unit 11b that creates a bounding box.

In addition, the animal semantic segmentation unit 12 includes a body semantic segmentation unit 12a that pixel-wise labels the animal's body, and an eye semantic segmentation unit 12b that pixel-wise labels the animal's eye region. It can be configured.

In this embodiment, the animal determination unit 22 determines that the number of pixels of the body generated by the body semantic segmentation unit 12a in the body area detection unit 11a is more than a predetermined number within a bounding box. Decide that animals exist.

In addition, the animal's eye position coordinate determination unit 13b is for determining the position of the animal's eyes, and the eye semantic segmentation unit 12b is located within the bounding box generated by the eye area detection unit 11b. If the number of pixels of the eyes created by occupies more than a predetermined number, it is determined that animal eyes exist within the bounding box.

If the animal's eye position coordinate determination unit 13b fails to determine the eye position, the eye position coordinates are determined based on the fitting similarity between the semantically segmented animal body region and the standard animal body with eye marks. Estimation is preferred.

Figure 3b is an example of the eye position coordinate determination unit 13 of an animal according to the present application.

The animal's eye position coordinate determination unit 13 determines the number of pixels of the animal's eyes 14b obtained by the eye semantic segmentation unit 12b within the bounding box 14a generated by the eye area detection unit 11b. It is determined whether the number is more than a predetermined number, and if it is more than a predetermined number, it is determined that there are animal eyes at the center of the bounding box, and this location becomes the target of laser attack. At this time, the laser beam of the laser generator 32a is directed to this location. is adjusted towards.

The position of the animal's eyes obtained by the eye position coordinate determination unit 13 is preferably displayed overlaid on the drone image and provided to the farm manager through the laser pointing screen 70b. In this case, the farm manager can press the laser firing button 70e on the mobile application 70 while checking the laser pointing screen 70b in real time.

Figure 3c shows several embodiments of a drone-based repelling attack means according to the present application.

Specifically, the eye laser attack of the drone 30, which fires a laser to the eyes, which are vulnerable body parts of the animal, through the laser generator 32a, and the drone 30 flies quickly up and down, producing an ultrasonic speaker 32b or sound. An example of a threatening flight of the drone 30 is shown, which threatens the animal by generating a loud sound or ultrasonic waves while approaching the animal using the speaker 32c.

The laser attack of the drone 30 is preferably performed to exterminate harmful animals by firing a laser into the animal's eyes through the laser generator 32a using the animal's appearance location and eye position coordinate information provided by the mobile device 40. do.

The threatening flight of the drone 30 accompanied by the sound or ultrasonic waves is performed after the drone 30 moves to the location coordinates specified in the animal location coordinate calculation unit 54 and then uses the ultrasonic speaker 32b or the acoustic speaker 32c. By using sound or ultrasonic waves to fly up and down quickly and threaten animals, they are used to exterminate harmful animals.

The ultrasonic speaker 32b is characterized in that it emits ultrasonic waves in a specified frequency band that responds sensitively to the type of animal identified by the animal determination unit 22.

For example, ultrasound in the specified frequency band is 13.5 kHz to 19.5 kHz for rats or wild dogs, 19.5 kHz to 24.5 kHz for raccoons or cats, 24.5 kHz to 45.5 kHz for rodents or birds, and 25 kHz to 30 kHz for elk. It is preferred to use high-bandwidth ultrasound.

The farm manager remotely activates the extermination attack means 32 on the drone 30 by wirelessly transmitting a drone control command to the drone 30 through the drone control command transmitter 23 using the mobile application 70 at a desired time. You can exterminate harmful animals by adjusting or controlling them.

However, in the laser attack, it is preferred that a laser is fired one at a time each time the farm manager checks the drone image in real time through the mobile application 70 and presses the laser firing button 70e on the mobile application 70.

The description of the present application described above is for illustrative purposes, and those skilled in the art will understand that the present application can be easily modified into other specific forms without changing its technical idea or essential features. Therefore, the embodiments described above should be understood in all respects as illustrative and not restrictive. For example, each component described as single may be implemented in a distributed manner, and similarly, components described as distributed may also be implemented in a combined form.

The scope of the present application is indicated by the claims described below rather than the detailed description above, and all changes or modified forms derived from the meaning and scope of the claims and their equivalent concepts should be construed as being included in the scope of the present application.

2 Farm thermal imaging camera
3 wireless transmitting unit
10 Animal Recognition Artificial Intelligence Neural Network
11 Animal area detection unit
11a body area detection unit
11b eye area detection unit
12 Animal semantic segmentation unit
12a Body semantic segmentation unit
12b Eye semantic segmentation unit
13 Eye position coordinate determination unit
14a Bounding box generated by eye area detection unit
14b Eye pixel labeled by eye semantic segmentation unit
20 control unit
22 Animal Judgment Department
23 Drone control command transmitter
30 drones
31 Drone thermal imaging camera
32 means of extermination attack
32a laser generator
32b ultrasonic speaker
33c acoustic speaker
40 mobile devices
52 farm grid map
54 Animal position coordinate calculation unit
70 mobile applications
70a real-time drone video
70b laser pointing screen
70c Attack method selection menu
70d sound selection menu
70e laser fire button
70f remote attack mode button
70g auto attack mode button

Claims (12)

In the AI-pest monitoring device using drones, including mobile applications installed on mobile devices and drones,
The drone includes a drone thermal imaging camera for photographing animals during aerial flight and a repelling attack means for repelling and attacking animals,
The mobile application includes an animal recognition artificial intelligence neural network for detecting the eyes and body of an animal exposed to the drone thermal imaging camera, and an attack means that can remotely select and activate the expulsion attack means through a touch screen on the mobile device. Containing a selection menu,
AI-harmful animal surveillance device using drones.
According to claim 1,
The repelling attack means includes one selected from the group consisting of a laser generator for laser attacking the eyes of the animal, an ultrasonic speaker, an acoustic speaker, and combinations thereof,
AI-harmful animal surveillance device using drones.
According to claim 1,
The mobile application includes a sound selection menu that allows you to select a sound type.
AI-harmful animal surveillance device using drones.
According to claim 1,
The mobile application includes a laser pointing screen that displays the position of the laser beam over the drone image,
While checking the laser pointing screen in real time, a laser beam is fired by pressing the laser firing button on the mobile application,
AI-harmful animal surveillance device using drones.
According to claim 1,
The animal recognition artificial intelligence neural network includes an animal region detection unit and an animal semantic segmentation unit,
The animal area detection unit includes a body area detection unit that generates a bounding box for the animal's body area and an eye area detection unit that generates a bounding box for the position of the animal's eyes,
The animal semantic segmentation unit includes a body semantic segmentation unit that pixels the body of the animal and an eye semantic segmentation unit that pixels the eye area of the animal.
AI-harmful animal surveillance device using drones.
According to claim 1,
The mobile application is
an eye position coordinate determination unit that determines the eye position of the animal from the results of the animal recognition artificial intelligence neural network; and
A drone control command transmitter for wirelessly transmitting a control command for the drone;
which further includes,
AI-harmful animal surveillance device using drones.
According to claim 1,
The AI-harmful animal monitoring device using the drone is
Farm thermal imaging cameras for filming animals; and
a wireless transmitter for providing images obtained by the farm thermal imaging camera to the mobile device;
which additionally includes,
AI-harmful animal surveillance device using drones.
According to claim 7,
The mobile application is
Comprising an animal position coordinate calculation unit for calculating the positions of animals detected on a farm grid map through images obtained from the farm thermal imaging camera and/or the drone thermal imaging camera, and specifying the position coordinates,
AI-harmful animal surveillance device using drones.
In AI-harmful animal surveillance device using drones,
The drone is,
Drone thermal imaging cameras for filming animals during mid-air flight;
Repelling attack means for repelling and attacking animals;
An animal recognition artificial intelligence neural network for detecting the eyes and body of an animal exposed to the drone thermal imaging camera;
An animal location coordinate calculation unit for specifying the location coordinates of animals detected on the farm grid map; and
an eye position coordinate determination unit that determines the eye position of the animal from the results of the animal recognition artificial intelligence neural network;
which includes,
AI-harmful animal surveillance device using drones.
According to clause 9,
It further includes a mobile application that receives drone images from the drone and provides them to the farm manager,
The farm manager controls the extermination attack means through remote control commands using buttons on the mobile application to exterminate the harmful animals,
AI-harmful animal surveillance device using drones.
In the AI-harmful animal monitoring device using a drone, which includes a drone and a remote server that remotely receives and analyzes drone images captured by the drone,
The remote server is,
An animal recognition artificial intelligence neural network for detecting the eyes and body of animals exposed to the drone image;
An animal location coordinate calculation unit for specifying the location coordinates of animals detected on the farm grid map; and
an eye position coordinate determination unit that determines the eye position of the animal from the results of the animal recognition artificial intelligence neural network;
Including,
The drone is,
Drone thermal imaging cameras for filming animals during mid-air flight; and
Repelling attack means for repelling and attacking animals;
Which includes,
AI-harmful animal surveillance device using drones.
According to claim 11,
It further includes a mobile application that receives drone images from the drone or remote server and provides them to the farm manager,
The farm manager controls the extermination attack means through a remote control command using a button on the mobile application to exterminate the harmful animals,
AI-harmful animal surveillance device using drones.