CN113743153A - Pet nose print identification method and system - Google Patents
Pet nose print identification method and system Download PDFInfo
- Publication number
- CN113743153A CN113743153A CN202010462705.0A CN202010462705A CN113743153A CN 113743153 A CN113743153 A CN 113743153A CN 202010462705 A CN202010462705 A CN 202010462705A CN 113743153 A CN113743153 A CN 113743153A
- Authority
- CN
- China
- Prior art keywords
- feature extraction
- layer
- pet
- convolutional
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
- Housing For Livestock And Birds (AREA)
Abstract
The disclosure relates to a pet nose print recognition method and system. According to an embodiment of the present disclosure, there is provided a pet nose print recognition method including inputting a nose print image of a pet into a convolutional neural network, the convolutional neural network including a feature extraction layer and a full connection layer, the full connection layer being connected after the feature extraction layer, the feature extraction layer including a plurality of feature extraction modules connected in series, each of the plurality of feature extraction modules including a pooling layer and a plurality of convolution layers, an input of at least one of the plurality of convolution layers of at least one of the plurality of feature extraction modules including an output of a previous convolution layer and an output of a convolution layer before the previous convolution layer, and the convolutional neural network being trained using a plurality of nose print images from different pets; based on the output of the convolutional neural network, a pet identity corresponding to the nose print image of the pet is determined.
Description
Technical Field
The present disclosure relates to the field of computer vision artificial intelligence, and more particularly to neural network-based pet nose print recognition.
Background
Image recognition refers to a technique of processing, analyzing and understanding an image with a computer to recognize various different patterns of objects and objects. It is an important field of artificial intelligence. In recent years, the ability of the deep learning model in the image classification field is exponentially improved, and the deep learning model becomes the most active research field in the artificial intelligence field. There is a need to apply image recognition techniques to more fields.
Disclosure of Invention
According to an embodiment of the present disclosure, there is provided a pet nose print recognition method, including: inputting a pet's nose print image into a convolutional neural network, wherein the convolutional neural network comprises a feature extraction layer and a fully connected layer, the fully connected layer is connected after the feature extraction layer, the feature extraction layer comprises a plurality of feature extraction modules connected in series, each feature extraction module of the plurality of feature extraction modules comprises a pooling layer and a plurality of convolutional layers, the input of at least one convolutional layer of the plurality of convolutional layers of at least one feature extraction module of the plurality of feature extraction modules comprises the output of a previous convolutional layer and the output of a convolutional layer before the previous convolutional layer, and wherein the convolutional neural network is trained using a plurality of nose print images from different pets; and determining a pet identity corresponding to the nose print image of the pet based on an output of the convolutional neural network.
The above summary of the solution is provided only to provide a basic understanding of various aspects of the subject matter described herein. Accordingly, the technical features in the above schemes are merely examples and should not be construed as limiting the scope or spirit of the subject matter described herein in any way. Other features, aspects, and advantages of the subject matter described herein will become apparent from the following detailed description, which, when taken in conjunction with the drawings.
Drawings
A better understanding of the present disclosure may be obtained when the following detailed description of the embodiments is considered in conjunction with the following drawings. The same or similar reference numbers are used throughout the drawings to refer to the same or like parts and operations. Wherein:
FIG. 1 shows a flow chart of a pet nose print recognition method according to one embodiment of the invention.
Fig. 2 shows a schematic diagram of an improved structure of the feature extraction module according to an embodiment of the present invention.
Fig. 3A and 3B show schematic diagrams of the outer product of convolution layers from the last two feature extraction modules of a plurality of feature extraction modules connected in series, according to an embodiment of the invention.
FIG. 4 illustrates a flow diagram for training the convolutional neural network using multiple rhinoprint images from different pets, according to one embodiment of the present invention.
Fig. 5 shows a schematic structural diagram of a bilinear convolutional neural network according to an embodiment of the present invention.
Detailed Description
Specific examples of aspects of the methods and systems according to the present disclosure are described below. These examples are described merely to add context and aid in the understanding of the described embodiments. It will thus be apparent to one skilled in the art that the embodiments described below may be practiced without some or all of the specific details. In other instances, well-known operations have not been described in detail so as not to unnecessarily obscure the described embodiments. Other applications are possible, and aspects of the present disclosure are not limited to these specific examples.
In real life, as the number of pets increases, the intelligent management of the pets becomes more and more important, the pets are lost and difficult to get back, and the wounded pets are difficult to trace the owners. More importantly, pet leash policies are difficult to enforce due to the health impact of pet chips, collars, and the like on individual pets. In order to solve the problems, the invention integrates pet nose print characteristic vectors of different levels through an improved neural network, and completes individual identity identification and authentication by utilizing subtle differences of pet nose prints. Convenient operation, pet friendliness, health and high intelligence degree.
According to one embodiment of the invention, a pet nose print recognition method is provided. As shown in fig. 1, the method includes step 101 and step 102.
In step 101, a pet's nose print image is input into a convolutional neural network, wherein the convolutional neural network comprises a feature extraction layer and a fully-connected layer, the fully-connected layer is connected after the feature extraction layer, the feature extraction layer comprises a plurality of feature extraction modules connected in series, each feature extraction module in the plurality of feature extraction modules comprises a pooling layer and a plurality of convolutional layers, the input of at least one convolutional layer in the plurality of convolutional layers of at least one feature extraction module in the plurality of feature extraction modules comprises the output of a previous convolutional layer and the output of a convolutional layer before the previous convolutional layer, and wherein the convolutional neural network is trained using a plurality of nose print images from different pets.
In step 102, a pet identity corresponding to a nose print image of the pet is determined based on an output of the convolutional neural network.
According to the method provided by the embodiment of the invention, the nose print of the individual pet can be identified to determine the unique identity of the individual pet, the method is used for intelligently monitoring the pet, the subjectivity of artificial identification is avoided, the identification accuracy is improved, and the influence of the traditional ear tag, the traditional neck ring and the like on the health of the pet is reduced. Scenarios to which the method according to one embodiment of the invention may be applied include: in the smart city pet management, pet certificates are transacted online, pets are lost and searched, pets are wounded and treated, a part of individual recognition scenes in smart agriculture, individual intelligent supervision, pet insurance, pet medical treatment and the like are realized.
The convolutional neural network used in the method according to one embodiment of the invention can fuse multi-level features, improve the accuracy of fine-grained identification and provide high-robustness nose print identification. The application scene of intelligent management of the pet is expanded, the influence of the traditional chip, the collar and the like on the individual health of the pet is reduced, and the intelligent management system is friendly to the pet, simple to operate and capable of checking quickly.
In one embodiment according to the invention, at least one feature extraction module of the plurality of feature extraction modules comprises a first convolutional layer, a second convolutional layer, a third convolutional layer, and a pooling layer connected in series, wherein an input of the third convolutional layer comprises an output of the second convolutional layer and an output of the first convolutional layer.
Fig. 2 shows a schematic diagram of an improved structure of the feature extraction module 201 according to an embodiment of the present invention. The feature extraction module 201 is one of a plurality of feature extraction modules of a convolutional neural network connected in series. Illustratively, the feature extraction module 201 may include a first convolution layer conv1_1, a second convolution layer conv1_2, a third convolution layer conv1_3, and a maximum pooling layer max _ pool. In one embodiment of the invention, the input of the third convolutional layer of the feature extraction module includes not only the output of the second convolutional layer but also the output of the first convolutional layer. That is, the output of the first convolutional layer conv1_1 and the output convolved by the second convolutional layer conv1_2 are simultaneously input to the nonlinear activation (e.g., Relu) of the third convolutional layer conv1_ 3. Thereby, a jump connection structure (in this example, a jump connection of the first convolution layer to the third convolution layer) in the feature extraction module is realized.
The same operations may also be implemented in other convolutional layers of the feature extraction module or may be implemented in other feature extraction modules. For example, a jump connection of the first convolution layer conv2_1 in the second feature extraction module to the third convolution layer conv2_3, a jump connection of the first convolution layer conv3_1 in the third feature extraction module to the third convolution layer conv3_3, a jump connection of the first convolution layer conv4_1 in the fourth feature extraction module to the third convolution layer conv4_3, a jump connection of the first convolution layer conv5_1 in the fifth feature extraction module to the third convolution layer conv5_3, and the like may be implemented. For another example, if the feature extraction module includes four convolutional layers, a skip connection of the first convolutional layer to the third convolutional layer, a skip connection of the first convolutional layer to the fourth convolutional layer, and/or a skip connection of the second convolutional layer to the fourth convolutional layer may be implemented. Those skilled in the art will appreciate that fig. 2 only schematically illustrates an embodiment in which feature fusion is achieved by means of a skip connection, and that skip connections of other convolutional layers may be implemented without departing from the technical solution of the present invention.
By adopting the jump connection, the convolutional neural network can fuse the features of different levels, and the fine-grained identification accuracy is improved, so that the convolutional neural network is particularly suitable for pet nose print identification.
In a method according to an embodiment of the invention, the input of the fully connected layer comprises the outer product of convolution layers from the last two feature extraction modules of the plurality of feature extraction modules connected in series.
Fig. 3A shows a schematic structural diagram of a convolutional neural network including the outer product of convolutional layers from the last two feature extraction modules of a plurality of feature extraction modules connected in series, according to one embodiment of the present invention.
As shown in fig. 3A, the feature extraction layer of the convolutional neural network may include five feature extraction modules connected in series. An outer product of convolution layers of a fourth feature extraction module and a fifth feature extraction module may be added to the convolutional neural network. Convolution layers may be selected from the fourth feature extraction module and the fifth feature extraction module, respectively, for example, a first convolution layer of the fourth feature extraction module and a third convolution layer of the fifth feature extraction module may be selected, a third convolution layer of the fourth feature extraction module and a third convolution layer of the fifth feature extraction module may be selected, an outer product is performed on the selected convolution layers, and an outer product result of the convolution layers is used as an input of the full connection layer.
Because the features extracted by the feature extraction modules at the positions, which are farther back, in the feature extraction layer are more valuable for classification, the outer product of the convolution layers of the fourth feature extraction module and the fifth feature extraction module is increased, so that the more valuable features at different levels can be fused, the fine-grained identification accuracy is improved, and the convolution neural network is particularly suitable for pet nose print identification.
According to one embodiment, one or more pairs of convolutional layers may be selected in the last two feature extraction modules, where the two convolutional layers of each pair are outer-multiplied, and the outer-product result of each pair of convolutional layers is used as the input of the fully-connected layer. The two convolution layers of each pair can come from different feature extraction modules, can also come from the same feature extraction module, and can even be the outer products of the same convolution layer and the convolution layer.
Fig. 3B shows a convolutional layer outer product schematic from the last two feature extraction modules of a plurality of feature extraction modules connected in series, according to one embodiment of the invention.
For example, as shown in fig. 3B, the features of different layers are sufficiently fused by performing an outer product on the first convolution layer conv4_1 of the fourth feature extraction module and the third convolution layer conv5_3 of the fifth feature extraction module, performing an outer product on the third convolution layer conv4_3 of the fourth feature extraction module and the third convolution layer conv5_3 of the fifth feature extraction module, performing an outer product on the third convolution layer conv5_3 of the fifth feature extraction module and itself, and inputting the final outer product result to the fully connected layer.
Those skilled in the art will appreciate that fig. 3A and 3B only schematically illustrate embodiments in which feature fusion is achieved by outer products, and that outer products of other convolutional layers may be achieved without departing from the technical solution of the present invention. For example, in an embodiment where the last two feature extraction modules respectively have four convolution layers, the third convolution layer and the fourth convolution layer of the last two feature extraction modules may be respectively selected to perform outer product two by two, and the outer product result may be input to the full-link layer for feature fusion.
By adopting the outer product, the convolutional neural network can fuse the features of different levels, and the fine-grained identification accuracy is improved, so that the convolutional neural network is particularly suitable for pet nose print identification.
FIG. 4 illustrates a flow diagram for training a convolutional neural network using multiple rhinoprint images from different pets, according to one embodiment of the present invention.
In step 401, a plurality of images of faces of different pets are obtained. In one embodiment, an image of the face of the pet may be captured by an image capture device. The image acquisition equipment can be mobile equipment with a camera shooting function, can automatically focus the face of the individual pet, and automatically supplements light according to environmental conditions. After the mobile device finishes collecting, the data can be transmitted to the server in real time by using an http protocol, for example. Training the convolutional neural network is performed at a server.
In one embodiment, the image acquisition equipment can automatically preview and identify parameters such as definition, brightness, exposure and the like of a shot image during shooting, and provide feedback when the image quality is unqualified. For example, a qualified image may include the texture of a full and clear pet nose, achieving picture quality that can train a convolutional neural network.
In another embodiment, a stored image of the pet's face may be obtained from a server.
In step 402, a plurality of images of the faces of the different pets are pre-processed.
According to one embodiment of the invention, the pre-processing may include adjusting at least one of sharpness, brightness, exposure of the image. For example, image pre-processing may include data cleansing, data enhancement, normalization, manual screening of images, and the like.
Training sets, validation sets, and test sets may also be partitioned for the images. For example, in one embodiment, the partition ratio may be a training set: and (4) verification set: test set 7:2: 1.
In step 403, a plurality of nose print images are obtained from the pre-processed plurality of images of the faces of the different pets. The nose pattern image may be obtained by clipping an outline of the nose of the pet in the image of the face of the pet based on the nose pattern feature point.
In step 404, feedback is provided when the image quality of a nose pattern image of the plurality of nose pattern images is not acceptable. For example, the image quality failing may be that the image quality of the nose print image cannot be used to train a convolutional neural network.
In step 405, corresponding pet identity labels are set for the plurality of nose print images respectively.
In step 406, the convolutional neural network is trained by using the plurality of nose print images and the corresponding pet identity labels, so that the trained convolutional neural network can identify the pet identity labels corresponding to the nose print images. In one embodiment, the output dimension of the classifier of the neural network is the number of the pet individuals, the loss function can be a classification loss function, and the parameter configuration of the convolutional neural network can be optimally adjusted by a random gradient descent algorithm by taking a cosine distance as a measurement mode.
In a method according to an embodiment of the invention, before training the convolutional neural network using a plurality of rhinoprint images from different pets, parameters of the convolutional neural network are initialized using model parameters obtained by training the convolutional neural network using open source large-scale data.
The convolutional neural network is trained, for example, by migration learning. And the transfer learning represents that the improved convolutional neural network is initialized by using model parameters based on open-source large-scale data training, and fine-grained identification is carried out by using the obtained data of the nose pattern image. Specifically, the existing model parameters are used as the starting points of the training model, so that the neural network can keep the skills learned in the old task and continuously learn. And accelerating the convergence speed of the new network on the one hand based on the migration learning. On the other hand, the influence of the excessively small data quantity on the scene recognition accuracy is reduced.
The trained neural network shown in fig. 4 may be used as, for example, the convolutional neural network mentioned in fig. 1, to identify the pet nose print. In the recognition process, similar to the training process, a nose print image of the pet may be obtained through a series of processes. According to an embodiment of the present invention, the pet nasal print recognition method shown in fig. 1 further includes obtaining a nasal print image of the pet, which includes: an image of a face of the pet is obtained. Preprocessing an image of the pet's face; obtaining a nose print image of the pet from the pre-processed image of the face of the pet; and providing feedback when the image quality of the pet's nose line image is not acceptable.
In one embodiment, an image of the face of the pet may be captured by an image capture device. The image acquisition equipment can be mobile equipment with a camera shooting function, can automatically focus the face of the individual pet, and automatically supplements light according to environmental conditions. After the mobile device finishes collecting, the data can be transmitted to the server in real time by using an http protocol, for example. The pet nose print recognition method is carried out on the server, and the server feeds back the recognition result to the mobile equipment.
In one embodiment, the image acquisition equipment can automatically preview and identify parameters such as definition, brightness, exposure and the like of a shot image during shooting, and provide feedback when the image quality is unqualified. For example, a qualified image may include the texture of a full and clear pet nose, achieving picture quality that can be identified using a convolutional neural network.
In another embodiment, a stored image of the pet's face may be obtained from a server.
Wherein the pre-processing may include adjusting at least one of sharpness, brightness, and exposure of the image. The nose pattern image may be obtained by clipping an outline of the nose of the pet in the image of the face of the pet based on the nose pattern feature point.
And inputting the preprocessed pictures to be recognized into the trained convolutional neural network, and carrying out forward propagation and classification on the convolutional neural network to obtain the identity labels corresponding to the pictures so as to realize the identity confirmation of the individual pets.
The convolutional neural network for identifying the pet nose print can only have one path of feature extraction layer, and also can have two paths of parallel feature extraction layers. Fig. 5 shows a schematic structural diagram of a bilinear convolutional neural network according to an embodiment of the present invention.
According to one embodiment of the present invention, the convolutional neural network as described above is a bilinear convolutional neural network. As shown in fig. 5, the bilinear convolutional neural network includes a full connection layer and two parallel feature extraction layers 501 and 502, where the full connection layer is connected behind the two parallel feature extraction layers, and the features extracted by the two parallel feature extraction layers are fused by the full connection layer. For example, each feature extraction layer may comprise five feature extraction modules 1-5 connected in series, each of which may have the structure of a feature extraction module such as described in connection with fig. 2.
Similar to the embodiment shown in fig. 3, the input to the fully-connected layer in fig. 5 may be the outer product of the convolutional layers of the last two feature extraction modules in each way of the feature extraction layer, i.e., the outer product of the convolutional layers in a total of four feature extraction modules. Unlike the embodiment of fig. 3, the convolution layers for outer product may be derived from not only the same feature extraction layer but also two feature extraction layers, respectively. For example, one or more pairs of convolutional layers may be selected from the four feature extraction modules, where two convolutional layers of each pair are outer-multiplied, and the outer-product result of each pair of convolutional layers is used as the input of the fully-connected layer. The two convolution layers of each pair can come from different feature extraction modules, can also come from the same feature extraction module, and can even be the outer products of the same convolution layer and the convolution layer. Structurally, as shown in fig. 5, since features are extracted by two parallel feature extraction layers and then multiplied by an outer product, a large number of pairwise combinations of features are generated. The feature extraction layers 501 and 502 can extract different features and then perform fusion operation on the features through the full connection layer. The improved bilinear network is used for fusing features of different layers and performing transfer learning, so that the training network model has better generalization. For example, the convolutional neural network may be a bilinear convolutional neural network obtained by modifying the original VGG16 network. The bilinear convolutional neural network can adopt the same VGG network, and is called as a symmetrical Two stream neural network.
Identity recognition can be carried out on the pet nose print image to be recognized based on the training model. The bilinear convolutional neural network reduces the memory, fully considers the spatial position relation of the features, fuses the features of different layers and is more suitable for identifying the pet nose print.
In one embodiment of the invention, the improved bilinear convolutional neural network is used for extracting and identifying the pet nose print image features, and compared with the prior art, the method has the advantages that multi-level features are fused, the identification accuracy is greatly improved, and the method is more suitable for identifying the fine granularity of the nose print. By means of the technology, the pet nose print images can be effectively classified and extracted, the labor and financial resources of pet management are saved, the management efficiency is improved, and the method is more robust in the actual individual intelligent supervision scene.
There is also provided, in accordance with an embodiment of the present invention, a pet nasal print recognition system, including: means for inputting a pet's nose print image into a convolutional neural network, wherein the convolutional neural network comprises a feature extraction layer and a fully connected layer, the fully connected layer being connected after the feature extraction layer, the feature extraction layer comprising a plurality of feature extraction modules connected in series, each feature extraction module of the plurality of feature extraction modules comprising a pooling layer and a plurality of convolutional layers, an input of at least one convolutional layer of the plurality of convolutional layers of at least one feature extraction module of the plurality of feature extraction modules comprising an output of a previous convolutional layer and an output of a convolutional layer before the previous convolutional layer, and wherein the convolutional neural network is trained using a plurality of nose print images from different pets; and means for determining a pet identity corresponding to the nasal print image of the pet based on an output of the convolutional neural network.
In a system according to an embodiment of the invention, at least one feature extraction module of the plurality of feature extraction modules comprises a first convolutional layer, a second convolutional layer, a third convolutional layer and a pooling layer connected in series, wherein an input of the third convolutional layer comprises an output of the second convolutional layer and an output of the first convolutional layer.
In a system according to an embodiment of the invention, the input of the fully connected layer further comprises the outer product of convolution layers from the last two feature extraction modules of the plurality of feature extraction modules connected in series.
In a system according to an embodiment of the invention, further comprising means for training the convolutional neural network using a plurality of rhinoprint images from different pets, comprising: means for obtaining a plurality of images of faces of different pets; means for pre-processing a plurality of images of the faces of the different pets; the pet identification method comprises the steps of obtaining a plurality of nose print images from a plurality of preprocessed images of the faces of different pets, providing feedback when the image quality of the nose print images in the nose print images is unqualified, setting corresponding pet identification labels for the nose print images respectively, and training the convolutional neural network by using the nose print images and the corresponding pet identification labels so that the trained convolutional neural network can identify the pet identification labels corresponding to the nose print images.
In a system according to an embodiment of the invention, further comprising means for obtaining a nasal print image of the pet, comprising: means for obtaining an image of a face of a pet; means for pre-processing an image of the pet's face; means for obtaining a nasal print image of the pet from the pre-processed image of the face of the pet; and means for providing feedback when the image quality of the pet's nose line image is not acceptable.
In the system according to an embodiment of the present invention, further comprising an image capturing device for capturing an image of the face of the pet, the image capturing device is capable of providing feedback when the captured image of the face of the pet is not qualified.
In the system according to an embodiment of the present invention, wherein the preprocessing includes adjusting at least one of sharpness, brightness, and exposure of the image, and the system further includes means for obtaining the nose line image by cutting an outline of the nose of the pet in the image of the face of the pet based on the nose line feature point.
In a system according to an embodiment of the invention, further comprising means for initializing parameters of the convolutional neural network using model parameters resulting from training the convolutional neural network with open source large scale data before training the convolutional neural network using a plurality of nasoprint images from different pets.
In the system according to an embodiment of the present invention, the convolutional neural network is a bilinear convolutional neural network, and the bilinear convolutional neural network includes a full connection layer and two parallel feature extraction layers, where the full connection layer is connected behind the two parallel feature extraction layers, and the features extracted by the two parallel feature extraction layers are fused by the full connection layer.
There is also provided, in accordance with an embodiment of the present invention, an electronic device including a processor and a memory having program instructions stored thereon that, when executed by the processor, perform the method as previously described.
There is also provided, in accordance with an embodiment of the present invention, a computer-readable storage medium having stored thereon computer instructions which, when executed by a processor, implement the method as previously described.
In some embodiments, memory may include installed media (e.g., CD-ROM, floppy disk, or tape devices), random access memory (such as DRAM, DDR RAM, SRAM, EDO RAM, Rambus RAM, etc.), non-volatile memory (such as flash memory, magnetic media, or optical storage), registers, or other similar types of memory elements, and so forth. Memory 1102 may also include other types of memory or combinations thereof.
The processor may be any processor that can be used to process information, such as a microprocessor, digital signal processor, microcontroller, multi-core processor, special purpose processor, interface for network communications, and the like. The processor may execute various software components stored in the memory device (as is possible according to embodiments of the present disclosure) to perform various functions of the system.
Aspects of the present disclosure may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects all of which may be referred to herein generally as a "circuit," module "or" system. Any combination of one or more computer-readable storage media may be used. The computer readable storage medium may be a computer readable signal medium or a computer readable storage medium.
A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
The present disclosure, in various embodiments, configurations, and aspects, includes components, methods, processes, systems and/or apparatus substantially as depicted and described herein, including various embodiments, subcombinations, and subsets thereof. Those of skill in the art will understand how to make and use the systems and methods disclosed herein after understanding the present disclosure. In various embodiments, configurations, and aspects, the disclosure includes providing apparatuses and processes in the absence of items not depicted and/or described herein or in various embodiments, configurations, or aspects hereof, including in the absence of items that may have been used in previous apparatuses or processes, e.g., for improving performance, achieving ease and/or reducing cost of implementation.
In addition, embodiments of the present disclosure may also include the following examples:
Item 2. the method of item 1, wherein at least one feature extraction module of the plurality of feature extraction modules comprises a first convolutional layer, a second convolutional layer, a third convolutional layer, and a pooling layer connected in series, wherein an input of the third convolutional layer comprises an output of the second convolutional layer and an output of the first convolutional layer.
Item 3. the method of item 1, wherein the input to the fully connected layer further comprises an outer product of convolution layers from the last two feature extraction modules of the plurality of feature extraction modules connected in series.
Item 4. the method of item 1, further comprising training the convolutional neural network using a plurality of rhinoprint images from different pets, comprising: obtaining a plurality of images of faces of different pets; preprocessing a plurality of images of the faces of the different pets; obtaining a plurality of nasal print images from a plurality of preprocessed images of the faces of different pets, providing feedback when the image quality of the nasal print images in the plurality of nasal print images is unqualified, respectively setting corresponding pet identity labels for the plurality of nasal print images, and training the convolutional neural network by using the plurality of nasal print images and the corresponding pet identity labels, so that the trained convolutional neural network can identify the pet identity labels corresponding to the nasal print images.
Item 5. the method of item 1, further comprising obtaining a nose print image of the pet, comprising: obtaining an image of a face of a pet; preprocessing an image of the pet's face; obtaining a nose print image of the pet from the pre-processed image of the face of the pet; and providing feedback when the image quality of the pet's nose line image is not acceptable.
Item 6. the method of item 4 or 5, wherein the image of the face of the pet is captured by an image capture device capable of providing feedback when the captured image of the face of the pet is not acceptable.
Item 7. the method of item 4 or 5, wherein the preprocessing comprises adjusting at least one of sharpness, brightness, and exposure of the image, and wherein the nose line image is obtained by clipping an outline of the nose of the pet in the image of the face of the pet based on the nose line feature points.
Item 8. the method of item 4, further comprising, prior to training the convolutional neural network using a plurality of rhinoprint images from different pets, initializing parameters of the convolutional neural network using model parameters resulting from training the convolutional neural network using open source large scale data.
Item 9. the method of item 1, wherein the convolutional neural network is a bilinear convolutional neural network comprising a fully-connected layer and parallel first and second feature extraction layers, wherein the fully-connected layer is connected after the parallel first and second feature extraction layers, the features extracted by the parallel first and second feature extraction layers are fused by the fully-connected layer, and wherein the input to the fully-connected layer comprises the outer product of the convolutional layer from the feature extraction module of the last two feature extraction modules of the first feature extraction layer and the convolutional layer from the feature extraction module of the last two feature extraction modules of the second feature extraction layer.
Item 10. a pet nose print recognition system, the system comprising: means for inputting a pet's nose print image into a convolutional neural network, wherein the convolutional neural network comprises a feature extraction layer and a fully connected layer, the fully connected layer being connected after the feature extraction layer, the feature extraction layer comprising a plurality of feature extraction modules connected in series, each feature extraction module of the plurality of feature extraction modules comprising a pooling layer and a plurality of convolutional layers, an input of at least one convolutional layer of the plurality of convolutional layers of at least one feature extraction module of the plurality of feature extraction modules comprising an output of a previous convolutional layer and an output of a convolutional layer before the previous convolutional layer, and wherein the convolutional neural network is trained using a plurality of nose print images from different pets; and means for determining a pet identity corresponding to the nasal print image of the pet based on an output of the convolutional neural network.
Item 11 the system of item 10, wherein at least one feature extraction module of the plurality of feature extraction modules comprises a first convolutional layer, a second convolutional layer, a third convolutional layer, and a pooling layer connected in series, wherein an input of the third convolutional layer comprises an output of the second convolutional layer and an output of the first convolutional layer.
Item 12. the system of item 10, wherein the input to the fully connected layer further comprises an outer product of convolution layers from the last two feature extraction modules of the plurality of feature extraction modules connected in series.
The system of item 10, further comprising means for training the convolutional neural network using a plurality of nose print images from different pets, comprising: means for obtaining a plurality of images of faces of different pets; means for pre-processing a plurality of images of the faces of the different pets; the pet identification method comprises the steps of obtaining a plurality of nose print images from a plurality of preprocessed images of the faces of different pets, providing feedback when the image quality of the nose print images in the nose print images is unqualified, setting corresponding pet identification labels for the nose print images respectively, and training the convolutional neural network by using the nose print images and the corresponding pet identification labels so that the trained convolutional neural network can identify the pet identification labels corresponding to the nose print images.
Item 14. the system of item 10, further comprising means for obtaining a nasal print image of the pet, comprising: means for obtaining an image of a face of a pet; means for pre-processing an image of the pet's face; means for obtaining a nasal print image of the pet from the pre-processed image of the face of the pet; and means for providing feedback when the image quality of the pet's nose line image is not acceptable.
Item 15. the system of item 13 or 14, further comprising an image capture device for capturing an image of the face of the pet, the image capture device capable of providing feedback when the captured image of the face of the pet is not acceptable.
Item 16. the system of item 13 or 14, wherein the pre-processing comprises adjusting at least one of a sharpness, a brightness, an exposure of the image, and the system further comprises means for obtaining the nose line image by clipping an outline of the nose of the pet in the image of the face of the pet based on the nose line feature points.
Item 17. the system of item 13, further comprising means for initializing parameters of the convolutional neural network using model parameters resulting from training the convolutional neural network with open source large scale data prior to training the convolutional neural network using a plurality of nasoprint images from different pets.
Item 18. the system of item 10, wherein the convolutional neural network is a bilinear convolutional neural network comprising a fully-connected layer and parallel first and second feature extraction layers, wherein the fully-connected layer is connected after the parallel first and second feature extraction layers, the features extracted by the parallel first and second feature extraction layers are fused by the fully-connected layer, and wherein the input to the fully-connected layer comprises the outer product of the convolutional layer from the feature extraction module of the last two feature extraction modules of the first feature extraction layer and the convolutional layer from the feature extraction module of the last two feature extraction modules of the second feature extraction layer.
Item 19. an electronic device comprising a processor and a memory having program instructions stored thereon that, when executed by the processor, perform the method of any of items 1-9.
Item 20. a computer readable storage medium, wherein the computer readable storage medium stores computer instructions that, when executed by a processor, implement the method of any of items 1-9.
Moreover, although the description of the present disclosure has included description of one or more embodiments, configurations, or aspects, certain variations and modifications, other variations, combinations, and modifications are within the scope of the disclosure, e.g., as may be within the skill and knowledge of those in the art, after understanding the present disclosure. The present disclosure is intended to obtain rights which include alternative embodiments, configurations, or aspects to the extent permitted, including alternate, interchangeable and/or equivalent structures, functions, ranges or steps to those claimed, whether or not such alternate, interchangeable and/or equivalent structures, functions, ranges or steps are specifically set forth herein. Nothing herein is intended to publicly dedicate any patentable technical solution.
Claims (20)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010462705.0A CN113743153A (en) | 2020-05-27 | 2020-05-27 | Pet nose print identification method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010462705.0A CN113743153A (en) | 2020-05-27 | 2020-05-27 | Pet nose print identification method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113743153A true CN113743153A (en) | 2021-12-03 |
Family
ID=78723581
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010462705.0A Pending CN113743153A (en) | 2020-05-27 | 2020-05-27 | Pet nose print identification method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113743153A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114283453A (en) * | 2021-12-28 | 2022-04-05 | 新瑞鹏宠物医疗集团有限公司 | Information acquisition method, device, storage medium and electronic device for stray animals |
| CN114332931A (en) * | 2021-12-29 | 2022-04-12 | 新瑞鹏宠物医疗集团有限公司 | Pet relation map updating method, device, equipment and computer-readable storage medium |
| CN114332938A (en) * | 2021-12-30 | 2022-04-12 | 新瑞鹏宠物医疗集团有限公司 | Pet nose print recognition management method and device, intelligent equipment and storage medium |
| CN114332932A (en) * | 2021-12-29 | 2022-04-12 | 新瑞鹏宠物医疗集团有限��司 | Health prediction method and device, electronic equipment and storage medium |
| CN115034912A (en) * | 2022-05-18 | 2022-09-09 | 新瑞鹏宠物医疗集团有限公司 | Pet insurance handling method and device |
| CN115146805A (en) * | 2022-05-19 | 2022-10-04 | 新瑞鹏宠物医疗集团有限公司 | Pet amusement park entering method based on pet nose print and related device |
| CN115689807A (en) * | 2022-10-14 | 2023-02-03 | 中国工商银行股份有限公司 | Smart hosting method, system, device, medium, and program product |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944458A (en) * | 2017-12-08 | 2018-04-20 | 北京维大成科技有限公司 | A kind of image-recognizing method and device based on convolutional neural networks |
| CN109948458A (en) * | 2019-02-25 | 2019-06-28 | 广东智媒云图科技股份有限公司 | Pet personal identification method, device, equipment and storage medium based on noseprint |
| US20190392202A1 (en) * | 2018-10-30 | 2019-12-26 | Baidu Online Network Technology (Beijing) Co., Ltd. | Expression recognition method, apparatus, electronic device, and storage medium |
| CN110647912A (en) * | 2019-08-15 | 2020-01-03 | 深圳久凌软件技术有限公司 | Fine-grained image recognition method and device, computer equipment and storage medium |
-
2020
- 2020-05-27 CN CN202010462705.0A patent/CN113743153A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107944458A (en) * | 2017-12-08 | 2018-04-20 | 北京维大成科技有限公司 | A kind of image-recognizing method and device based on convolutional neural networks |
| US20190392202A1 (en) * | 2018-10-30 | 2019-12-26 | Baidu Online Network Technology (Beijing) Co., Ltd. | Expression recognition method, apparatus, electronic device, and storage medium |
| CN109948458A (en) * | 2019-02-25 | 2019-06-28 | 广东智媒云图科技股份有限公司 | Pet personal identification method, device, equipment and storage medium based on noseprint |
| CN110647912A (en) * | 2019-08-15 | 2020-01-03 | 深圳久凌软件技术有限公司 | Fine-grained image recognition method and device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| TSUNG-YU LIN 等: "Bilinear CNN Models for Fine-grained Visual Recognition", 《ARXIV:1504.07889V1》, 29 April 2015 (2015-04-29), pages 1 - 11 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114283453A (en) * | 2021-12-28 | 2022-04-05 | 新瑞鹏宠物医疗集团有限公司 | Information acquisition method, device, storage medium and electronic device for stray animals |
| CN114332931A (en) * | 2021-12-29 | 2022-04-12 | 新瑞鹏宠物医疗集团有限公司 | Pet relation map updating method, device, equipment and computer-readable storage medium |
| CN114332932A (en) * | 2021-12-29 | 2022-04-12 | 新瑞鹏宠物医疗集团有限公司 | Health prediction method and device, electronic equipment and storage medium |
| CN114332938A (en) * | 2021-12-30 | 2022-04-12 | 新瑞鹏宠物医疗集团有限公司 | Pet nose print recognition management method and device, intelligent equipment and storage medium |
| CN115034912A (en) * | 2022-05-18 | 2022-09-09 | 新瑞鹏宠物医疗集团有限公司 | Pet insurance handling method and device |
| CN115146805A (en) * | 2022-05-19 | 2022-10-04 | 新瑞鹏宠物医疗集团有限公司 | Pet amusement park entering method based on pet nose print and related device |
| CN115689807A (en) * | 2022-10-14 | 2023-02-03 | 中国工商银行股份有限公司 | Smart hosting method, system, device, medium, and program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113743153A (en) | Pet nose print identification method and system | |
| Fujita et al. | Basic investigation on a robust and practical plant diagnostic system | |
| Tariq et al. | Detecting both machine and human created fake face images in the wild | |
| Do et al. | Forensics face detection from GANs using convolutional neural network | |
| CN109614907B (en) | Pedestrian re-identification method and device based on feature-enhanced guided convolutional neural network | |
| Huh et al. | Fighting fake news: Image splice detection via learned self-consistency | |
| CN113505682B (en) | Living body detection method and living body detection device | |
| CN110321870B (en) | Palm vein identification method based on LSTM | |
| CN112347964B (en) | Behavior detection method and device based on graph network | |
| CN111598065B (en) | Depth image acquisition method and living body recognition method, device, circuit and medium | |
| CN115880558A (en) | Farming behavior detection method, device, electronic equipment and storage medium | |
| KR20210051473A (en) | Apparatus and method for recognizing video contents | |
| CN113807237B (en) | Training of in vivo detection model, in vivo detection method, computer device, and medium | |
| CN112199976A (en) | Credential image generation method and device | |
| CN109344700A (en) | A Pedestrian Pose Attribute Recognition Method Based on Deep Neural Network | |
| CN114038045A (en) | Cross-modal face recognition model construction method and device and electronic equipment | |
| CN111291773A (en) | Feature identification method and device | |
| CN115937596A (en) | Target detection method and its model training method, device and storage medium | |
| CN111079930A (en) | Method, device and electronic device for determining quality parameter of data set | |
| CN114519863A (en) | Human body weight recognition method, human body weight recognition apparatus, computer device, and medium | |
| CN115578673A (en) | Certificate authenticity verification method, system, computer device and storage medium | |
| CN112101395A (en) | Image identification method and device | |
| CN114429635A (en) | Book management method | |
| CN113591893B (en) | Image processing method and device based on artificial intelligence and computer equipment | |
| CN109522861A (en) | A kind of micro- expression recognition method of face multiclass |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211203 |
|
| RJ01 | Rejection of invention patent application after publication |