KR20200094840A - Animal Recognition Apparatus And Animal Recognition Method - Google Patents

Abstract

The present invention relates to an animal recognition device and to an animal recognition method. The animal recognition device comprises: a first imaging unit photographing a first image; a second imaging unit photographing a second image; and a control unit receiving images photographed by the first imaging unit and the second imaging unit, extracting an eye area image and an area around the eye from each image, and recognizing a specific animal by comparing the extracted images with pre-stored images.

Description

Animal Recognition Apparatus And Animal Recognition Method {Animal Recognition Apparatus And Animal Recognition Method}

The present invention relates to an animal recognition device and an animal recognition method, and more particularly, to an animal recognition device and an animal that captures an image using two photographing devices and recognizes a specific animal through comparison of images extracted from the captured image. It's about how to recognize.

After the mandatory enforcement of the animal registration system, the incidence of animal abandonment has rather increased, and the companion animal linkage project is also not active due to the low animal registration rate. Currently, the registration method of the animal registration system is performed using a biograft chip, an external necklace, and an identification tag.

The use of a biotransplantation chip has many advantages in its use. Nevertheless, there are several side effects to the use of the biotransplantation chip, so caregivers are hesitant to use it.

Therefore, there is a need to improve the animal registration method more efficiently and simply without harming the animals.

[Patent Document 1] Korean Patent Registration No. 10-1862639. 2018.05.24. Enrollment.

The present invention has been created to solve the above-described problems, and an object thereof is to provide an animal recognition apparatus and an animal recognition method.

The objects of the present invention are not limited to the above-mentioned objects, and other objects not mentioned will be clearly understood from the following description.

In order to achieve the above objects, an animal recognition apparatus and an animal recognition method according to an embodiment of the present invention are disclosed. The device receives a first image pickup unit for capturing a first image, a second image capturing unit for capturing a second image, and an image captured by the first and second image capturing units, and the eye part image from each image. And a control unit that extracts an image around the eye and compares the extracted images with pre-stored images to recognize a specific animal.

In addition, according to an embodiment of the present disclosure, a sound output unit for generating and outputting learned sound to fix the gaze and motion of an animal to be photographed and an image output unit for outputting a result recognized by the control unit may be further included. have.

In addition, according to an embodiment of the present disclosure, the learned sound may be at least one of a clicker sound or a recorded sound.

In addition, according to an embodiment of the present disclosure, the first imaging unit may be an infrared imaging device, and the second imaging unit may be a visible light imaging device.

In addition, according to an embodiment of the present disclosure, the controller may extract the eye portion image and the eye peripheral portion image by inputting the first image and the second image into a multilayer perceptron model. .

In addition, according to an embodiment of the present disclosure, the controller inputs the extracted eye portion image and the eye peripheral portion image into a deep learning-based convolutional neural network, and the iris image and the eye peripheral feature portion image, respectively. An image preprocessor and an image preprocessor that generates one or more, normalizes the size of the generated iris image and peripheral feature partial image, and encodes the normalized iris image and peripheral feature partial image, and the encoded iris image and peripheral feature partial image. It may include an animal recognition unit for recognizing a specific animal by comparing pre-stored images.

In addition, according to an embodiment of the present disclosure, the iris image may be an image characterized by either a shape of an iris or a color of an iris, and the characteristic partial image around the eye may include a shape of an eyelid and a shape of hair around the eye. Alternatively, it may be an image characterized by any one of the colors of the hair around the eyes.

Additionally, according to an embodiment of the present disclosure, the animal recognition method includes a sound output step of generating and outputting a learned sound to fix the gaze and motion of an animal to be photographed, and output of the learned sound in the sound output step. When this is completed, the first image and the second image respectively captured by the first and second imaging units are input, the first image and the second image input through the image input step, respectively. And an image extraction step of extracting an image around the eye, an image comparison and recognition step of recognizing a specific animal by comparing the extracted images with pre-stored images, and an image output step of outputting a result of recognition in the image comparison and recognition step. I can.

In addition, according to an embodiment of the present disclosure, the learned sound may be at least one of a clicker sound or a recorded sound.

In addition, according to an embodiment of the present disclosure, the first imaging unit may be an infrared imaging device, and the second imaging unit may be a visible light imaging device.

In addition, according to an embodiment of the present disclosure, in the step of extracting an image, the first image and the second image are input to a multilayer perceptron model to extract the eye part image and the eye area image. It can be a step.

In addition, according to an embodiment of the present disclosure, in the image comparison and recognition step, the extracted eye portion image and the eye peripheral portion image are input to a deep learning-based convolutional neural network, respectively, and An image pre-processing step of generating one or more feature partial images, normalizing the size of the generated iris image and eye peripheral feature partial image, and encoding the normalized iris image and eye peripheral feature partial image, and the encoded iris image and eye peripheral features It may include an animal recognition step of recognizing a specific animal by comparing the partial images with pre-stored images.

In addition, according to an embodiment of the present disclosure, the iris image may be an image characterized by either a shape of an iris or a color of an iris, and the characteristic partial image around the eye may include a shape of an eyelid and a shape of hair around the eye. Alternatively, it may be an image characterized by any one of the colors of the hair around the eyes.

Specific details for achieving the above objects will be clarified with reference to embodiments to be described later in detail with the accompanying drawings.

However, the present invention is not limited to the embodiments disclosed below, and may be configured in various different forms, to make the disclosure of the present invention complete, and to those skilled in the art to which the present invention pertains ( Hereinafter, it is provided to completely inform the scope of the invention to "normal engineer").

The present invention relates to an animal recognition apparatus and an animal recognition method for capturing an image using two photographing devices and recognizing a specific animal through the captured image.

In the present invention, an image is captured using two photographing devices, and an iris pattern can be accurately obtained by analyzing the photographed image.

In addition, the present invention can improve the iris recognition rate by applying a deep learning-based convolutional neural network.

In addition, the present invention can fix the animal's gaze and movement using the learned sound, thereby alleviating the fear of animals that fear the camera.

Therefore, according to the animal recognition apparatus disclosed in the present invention, it is possible to increase the registration rate of the animal by recognizing the animal using the iris of the animal.

The effects of the present invention are not limited to the above-described effects, and the potential effects expected by the technical features of the present invention will be clearly understood from the following description.

Some of the embodiments are shown in the accompanying drawings so that the above-mentioned features of the present invention can be understood in detail, with a more specific description, with reference to the following embodiments. In addition, similar reference numerals with the drawings are intended to refer to the same or similar functions across various aspects. It should be noted, however, that the accompanying drawings show only specific exemplary embodiments of the present invention, and are not considered to limit the scope of the present invention, and other embodiments having the same effect can be sufficiently recognized. Do it.
1 is a view showing the components of an animal recognition apparatus according to an embodiment of the present invention.
2 is a flowchart illustrating a method for recognizing an animal according to an embodiment of the present invention.

The present invention can be applied to various changes, and may have various embodiments, and specific embodiments will be illustrated in the drawings and described in detail.

Various features of the invention disclosed in the claims may be better understood in consideration of the drawings and detailed description. The devices, methods, manufacturing methods, and various embodiments disclosed in the specification are provided for illustrative purposes. The disclosed structural and functional features are intended to enable those skilled in the art to implement various embodiments in detail, and not to limit the scope of the invention. The terminology and sentences disclosed are intended to facilitate understanding of the various features of the disclosed invention and are not intended to limit the scope of the invention.

In describing the present invention, when it is determined that detailed descriptions of related known technologies may unnecessarily obscure the subject matter of the present invention, detailed descriptions thereof will be omitted.

Hereinafter, an animal recognition apparatus and an animal recognition method according to an embodiment of the present invention will be described.

1 is a view showing the components of an animal recognition apparatus according to an embodiment of the present invention.

Referring to FIG. 1, the animal recognition apparatus 100 of the present invention may include a sound output unit 110, an image capture unit 130, a control unit 150, and an image output unit 170.

In an embodiment, the sound output unit 110 may be provided in the animal recognition apparatus 100 to generate and output a learned sound that fixes the gaze and motion of an animal to be photographed. For example, animals, unlike humans, have difficulty consciously fixing their gaze. So, the learned sound can be used to induce the animal's gaze and motion fixation. The learned sound may include at least one of a clicker sound or a recorded sound. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the image pickup unit 130 includes an eye image and an image around the eye through a first image pickup unit for capturing a first image and a second image capturing unit for capturing a second image, that is, two image capturing devices. You can shoot a video More specifically, in the image capture unit 130, the first imaging unit may be an infrared imaging device, and the second imaging unit may be a visible light imaging device. The infrared imaging device, which is the first imaging unit, may capture an image including the pupil and iris locations of the eye. The visible ray photographing apparatus, which is the second imaging unit, may photograph an image including an image around the eyes.

The iris shows the best discrimination power among biometric information because each individual has its own characteristics. In addition, since registration and recognition are possible in a non-contact manner, a method of identifying an individual using an iris may be an effective method.

According to the present invention, the iris and eyeball periphery features, which have unique features for each individual, are synthesized and used for individual identification, thereby increasing the completeness of individual identification. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the control unit 150 receives the first image and the second image captured through the image pickup unit 130, extracts the eye part image and the eye area image from each image, and extracts A specific animal can be recognized by comparing images and previously stored images. The controller 150 may input the first image and the second image to a multilayer perceptron model to extract an eye portion image and an eye peripheral portion image.

More specifically, the control unit 150 may include an image preprocessor and an animal recognition unit. The image preprocessor inputs the eye part image and the eye part image extracted through the multilayer perceptron model into a deep learning-based convolutional neural network, and the iris image and the feature part image around the eye, respectively. One or more can be generated, the sizes of the generated iris image and the feature partial image around the eye can be normalized, and the normalized iris image and the feature partial image around the eye can be encoded. The animal recognition unit may recognize a specific animal by comparing the encoded iris image and the feature portion image around the eye with pre-stored images.

The iris image may be an image characterized by either the shape of the iris or the color of the iris, and the partial image of features around the eye is characterized by any one of the shape of the eyelid, the shape of the hair around the eye, or the color of the hair around the eye. It can be an image. Normalization can be used to make the size of the input image constant and to minimize the inclusion of redundant information. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In addition, artificial intelligence is a field of computer engineering and information technology that studies how computers can do the thinking, learning, and self-development that can be done with human intelligence. Means to be able to.

The controller 150 of the present invention may extract an eye portion image and an eye peripheral portion image from a captured image through a multilayer perceptron (MLP) model. The artificial neural network according to an embodiment of the present invention includes a convolution neural network (CNN), an auto encoder, a feedforward neural network, a radial basis function network, and a Cohen self-organization. A neural network (kohonen self-organizing network), a recurrent neural network (RNN) and a multilayer perception network (Multilayer Perceptron: MLP) may be used, and the present invention is not limited thereto and may include any neural network.

The pre-determined recognition algorithm refers to pre-trained to extract an iris image and an image around the eye using deep learning of artificial intelligence. More specifically, the iris image and the image around the eye are recognized using a confidence map, each iris image and the image around the eye are correlated using PAF (Part Affinity Fields), and the iris image and the eye area around the eye using the Greedy algorithm. You can analyze the video. The confidence map refers to a function of dividing and displaying organizational elasticity data according to reliability. The PAF (Part Affinity Fields) refers to a method of associating body parts with individuals in an image using a non-parameter. The Greedy algorithm refers to an algorithm that solves the entire problem while making greedy choices every moment when solving the entire problem. Therefore, the present invention learns tens of thousands of images using artificial intelligence deep learning technology and analyzes the user's iris image and images around the eyes using the above-described method, and thus the iris image and the image around the eyes without additional additional devices such as Kinect. Analyze and give feedback.

For example, when a dog's face is photographed through a first image pickup unit and a second image pickup unit, the photographed first image and the second image are input to a multilayer recognition network (MLP) to provide an image of the dog's eyes and around the eyes. Can be extracted. In addition, an iris image and a feature portion image of the eye area may be extracted by inputting the extracted eye and eye peripheral image into a convolutional neural network model. The extracted iris image and eye peripheral feature image are normalized and encoded. Then, it is possible to recognize that the animal is a specific animal by comparing the encoded iris image and the feature portion image around the eye with the previously stored images. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In an embodiment, the image output unit 170 may output a result recognized by the control unit 150. For example, the recognition result may include the photographed picture, the name of the animal, the individual number, the iris code, the breed, the date of birth, the registration area, and the caregiver information. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

2 is a flowchart illustrating a method for recognizing an animal according to an embodiment of the present invention.

Referring to Figure 2, the animal recognition method (S200) of the present invention is a sound output step (S210) of generating and outputting a learned sound to fix the gaze and motion of an animal to be photographed, in the sound output step, the learned When the output of the sound is completed, the first image and the second image captured by the first and second imaging units are respectively input (S230), the first image and the second image input through the image input step. In the image extraction step (S250) of extracting the eye part image and the eye area image from the image, the image comparison and recognition step (S270) of recognizing a specific animal by comparing the extracted images with pre-stored images, and the image comparison and recognition step An image output step S290 of outputting the recognized result may be included.

In an embodiment, the animal recognition method S200 may include a sound output step S210 of generating and outputting a learned sound to fix the gaze and motion of an animal to be photographed. For example, animals, unlike humans, have difficulty consciously fixing their gaze. So, the learned sound can be used to induce the animal's gaze and motion fixation. The learned sound may include at least one of a clicker sound or a recorded voice. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the animal recognition method (S200) is, in the sound output step, when the output of the learned sound is completed, the first image and the second image respectively captured by the first and second imaging units are inputted. It may include an input step (S230). The first image captured by the first imaging unit may be an image including an eye area image, and the second image captured by the second imaging unit may be an image including an eye area image.

More specifically, the image input step S230 may include capturing a first image and a second image, respectively, using an infrared imaging device as a first imager and a visible light imager as a second imager. The infrared imaging device, which is the first imaging unit, may capture an image including the pupil and iris locations of the eye. The visible ray photographing apparatus, which is the second imaging unit, may photograph an image including an image around the eyes.

The iris shows the most outstanding identification among biometric information because each individual has its own characteristics, and it can be registered and recognized in a non-contact manner, thereby enabling more effective use.

The iris image may be an image characterized by either the shape of the iris or the color of the iris, and the partial image of features around the eye is characterized by any one of the shape of the eyelid, the shape of the hair around the eye, or the color of the hair around the eye. It can be an image. According to the present invention, the iris and eyeball periphery features, which have unique features for each individual, are synthesized and used for individual identification, thereby increasing the completeness of individual identification. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the animal recognition method (S200) includes an image extraction step (S250) of extracting an eye part image and a peripheral part image from the first image and the second image input through the image input step (S230), respectively. can do.

In addition, the image extraction step S250 may be a step of extracting an eye portion image and an eye peripheral portion image by inputting the first image and the second image into a multilayer perceptron model.

More specifically, the image extraction step S250 may extract an eye portion image and an eye peripheral portion image from an image photographed through a multilayer perceptron model. The artificial neural network according to an embodiment of the present invention includes a convolution neural network (CNN), an auto encoder, a feedforward neural network, a radial basis function network, and a Cohen self-organization. A neural network (kohonen self-organizing network), a recurrent neural network (RNN), a multilayer perception network (Multilayer Perceptron), etc. may be configured, and the present invention is not limited thereto, and may include any neural network. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In one embodiment, the animal recognition method S200 may include an image comparison and recognition step S270 of recognizing a specific animal by comparing extracted images and pre-stored images.

More specifically, the image comparison and recognition step S270 may include an image preprocessing step and an animal recognition step. In the image pre-processing step, the eye part image and the eye part image extracted through the multilayer perceptron model are input into a deep learning-based convolutional neural network, and the iris image and the peripheral feature part are respectively. It may be a step of generating one or more images, normalizing the size of the generated iris image and the feature partial image around the eye, and encoding the normalized iris image and the feature partial image around the eye. The animal recognition step may be a step of recognizing a specific animal by comparing pre-stored images with an encoded iris image and an eye area feature partial image.

The iris image may be an image characterized by either the shape of the iris or the color of the iris, and the partial image of features around the eye is characterized by any one of the shape of the eyelid, the shape of the hair around the eye, or the color of the hair around the eye. It can be an image. Normalization can be used to make the size of the input image constant and to minimize the inclusion of redundant information. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

In an embodiment, the animal recognition method S200 may include an image output step S290 of outputting a result recognized in the image comparison and recognition step S270. For example, the analysis result may include a photographed picture, an animal's name, an individual number, an iris code, a breed, a date of birth, registration area, and caregiver information. The above example is only an example for explaining the present disclosure, and the present disclosure is not limited thereto.

The above description is merely illustrative of the technical spirit of the present invention, and those skilled in the art will be able to make various changes and modifications without departing from the essential characteristics of the present invention.

Therefore, the embodiments disclosed in this specification are not intended to limit the technical spirit of the present invention, but to illustrate, and the scope of the present invention is not limited by these embodiments.

The scope of protection of the present invention should be interpreted by the claims, and it should be understood that all technical spirits within the scope equivalent thereto are included in the scope of the present invention.

Claims (13)

In the animal recognition device,
A first imaging unit for photographing a first image;
A second imaging unit for photographing a second image; And
Recognizes a specific animal by receiving an image captured by the first and second imaging units, extracting an eye area image and an eye area image from each image, and comparing the extracted images with pre-stored images A control unit;
Containing,
Animal recognition device.
The method of claim 1,
A sound output unit for generating and outputting a learned sound that fixes the gaze and motion of an animal to be photographed; And
An image output unit outputting a result recognized by the control unit;
Further comprising,
Animal recognition device.
According to claim 2,
The learned sound is,
At least one of a clicker sound or a recorded voice,
Animal recognition device.
The method of claim 1,
The first imaging unit is an infrared imaging device,
The second imaging unit is a visible light imaging device,
Animal recognition device.
The method of claim 1,
The control unit,
Inputting the first image and the second image into a multilayer perceptron model to extract the eye part image and the eye area image,
Animal recognition device.
The method of claim 5,
The control unit,
By inputting the extracted eye portion image and eye peripheral portion image into a deep learning-based convolutional neural network model, each iris image and one or more eye peripheral feature portions are generated, and the generated iris image and eye An image preprocessor for normalizing the size of the peripheral feature partial image and encoding the normalized iris image and the eye peripheral feature partial image; And
An animal recognition unit for recognizing a specific animal by comparing the encoded iris image and the feature portion image around the eye with pre-stored images;
Containing,
Animal recognition device.
The method of claim 6,
The iris image,
An image characterized by either the shape of the iris or the color of the iris,
The partial image of the features around the eyes,
An image characterized by any one of the shape of the eyelid, the shape of the hair around the eyes, or the color of the hair around the eyes,
Animal recognition device.
In the animal recognition method,
A sound output step of generating and outputting a learned sound for fixing the gaze and motion of an animal to be photographed;
An image input step of inputting a first image and a second image respectively photographed by a first image pickup unit and a second image pickup unit when output of the learned sound is completed in the sound output step;
An image extraction step of extracting an eye portion image and an eye peripheral portion image from the first image and the second image input through the image input step, respectively;
An image comparison and recognition step of recognizing a specific animal by comparing the extracted images and pre-stored images; And
An image output step of outputting a result recognized in the image comparison and recognition step;
Containing,
Animal recognition method.
The method of claim 8,
The learned sound is,
At least one of a clicker sound or a recorded voice,
Animal recognition method.
The method of claim 8,
The first imaging unit is an infrared imaging device,
The second imaging unit is a visible light imaging device,
Animal recognition method.
The method of claim 8,
The image extraction step,
The step of extracting the eye part image and the eye area image by inputting the first image and the second image into a multilayer perceptron model,
Animal recognition method.
The method of claim 8,
The image comparison and recognition step,
By inputting the extracted eye portion image and eye peripheral portion image into a deep learning-based convolutional neural network model, each iris image and one or more eye peripheral feature portions are generated, and the generated iris image and eye An image preprocessing step of normalizing the size of the peripheral feature partial image and encoding the normalized iris image and the eye peripheral feature partial image; And
An animal recognition step of recognizing a specific animal by comparing pre-stored images with the encoded iris image and the feature partial image around the eye;
Containing,
Animal recognition method.
The method of claim 12,
The iris image,
It is an image characterized by either the shape of the iris or the color of the iris,
The partial image of the features around the eyes,
An image characterized by any one of the shape of the eyelid, the shape of the hair around the eyes, or the color of the hair around the eyes,
Animal recognition method.

Images