Non-contact fish individual identification method and system based on iris image

Abstract

The application discloses a non-contact fish individual identification method and system based on an iris image, and the method comprises the following steps: acquiring an image of a fish to be identified; performing iris segmentation on the image of the fish to be identified to obtain an iris image; performing normalization processing on the iris image, and performing texture feature extraction on the normalized iris image; inputting the texture features into three classifiers respectively, performing adaptive weighted voting fusion on the identification results of the three classifiers, and obtaining the identification result of the fish individual; when the identification results of the three classifiers are consistent, directly outputting the identification result of the fish individual; when the identification results of the three classifiers are inconsistent, multiplying the identification probability value of each classifier by the weight coefficient of the corresponding classifier to obtain the final probability value of each classifier, selecting the identification result of the classifier with the maximum final probability value as the final identification result, and outputting the final identification result.

Description

Technical Field

[0001] This invention relates to the field of fish individual identification technology, and in particular to a non-contact fish individual identification method and system based on iris images. Background Technology

[0002] The statements in this section merely refer to the background art related to this invention and do not necessarily constitute prior art.

[0003] The aquaculture industry is currently undergoing a transformation from traditional extensive farming models to modern intensive and precision farming models. In the farming, transportation, and sales of high-quality, economically valuable, or ornamental fish, due to their high individual value, long farming cycles, and significant impact on the industry, the identification of individual fish and the storage, transmission, management, and traceability of their physiological, physical, farming process, and other attribute information are of paramount importance. Monitoring changes in fish growth curves in marine ranches, studying behavioral patterns, identifying individual differences, and understanding fish community migration all rely on individual identification.

[0004] Currently, the most common method for individual fish identification relies on tools to label and measure fish body parameters. This method is invasive, and invasive methods can lead to technical, health, and behavioral problems. Even currently available non-invasive methods (e.g., external colorants) can cause behavioral changes and pose health risks. Furthermore, invasive identification is time-consuming and incurs significant costs. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a non-contact fish individual identification method and system based on iris images; it utilizes the unique biological identity information of each fish to achieve accurate identification of individual fish, and the image acquisition process is non-invasive and contactless.

[0006] On the one hand, a non-contact fish individual identification method based on iris images is provided;

[0007] Non-contact fish identification methods based on iris images include:

[0008] Acquire an image of the fish to be identified;

[0009] Iris segmentation is performed on the image of the fish to be identified to obtain an iris image;

[0010] The iris image is normalized, and texture features are extracted from the normalized iris image.

[0011] Texture features are input into three classifiers respectively. The recognition results of the three classifiers are adaptively weighted and fused to obtain the recognition result of the individual fish: when the recognition results of the three classifiers are consistent, the recognition result of the individual fish is directly output; when the recognition results of the three classifiers are inconsistent, the recognition probability value of each classifier is multiplied by the corresponding classifier weight coefficient to obtain the final probability value of each classifier. The recognition result of the classifier with the highest final probability value is selected as the final recognition result.

[0012] On the other hand, a non-contact fish individual identification system based on iris images is provided;

[0013] A non-contact fish identification system based on iris images includes:

[0014] The acquisition module is configured to acquire an image of the fish to be identified.

[0015] The iris segmentation module is configured to perform iris segmentation on the image of the fish to be identified, thereby obtaining an iris image;

[0016] The feature extraction module is configured to: normalize the iris image and extract texture features from the normalized iris image;

[0017] The weighted fusion module is configured to: input texture features into three classifiers respectively, perform adaptive weighted voting fusion on the recognition results of the three classifiers to obtain the recognition result of the individual fish; when the recognition results of the three classifiers are consistent, the recognition result of the individual fish is directly output; when the recognition results of the three classifiers are inconsistent, the final probability value of each classifier is obtained by multiplying the recognition probability value of each classifier by the corresponding weight coefficient, and the recognition result of the classifier with the largest final probability value is selected as the final recognition result output.

[0018] Furthermore, an electronic device is also provided, including:

[0019] Memory, used for non-transitory storage of computer-readable instructions; and

[0020] Processor, for executing the computer-readable instructions,

[0021] When the computer-readable instructions are executed by the processor, they perform the method described in the first aspect above.

[0022] In another aspect, a storage medium is also provided for non-transitory storage of computer-readable instructions, wherein when the non-transitory computer-readable instructions are executed by a computer, the instructions of the method described in the first aspect are executed.

[0023] In another aspect, a computer program product is also provided, including a computer program that, when run on one or more processors, is used to implement the method described in the first aspect above.

[0024] One of the above technical solutions has the following advantages or beneficial effects:

[0025] This method enables individual identification of different individuals of the same fish species based on fish-eye iris recognition. It includes an image acquisition device for capturing iris images of individual fish, iris segmentation, and individual fish identification. Individual fish identification involves normalizing the segmented iris images, extracting texture features, and classifying features. Using the image acquisition device, a telecentric lens is used to capture images of fish swimming through a connecting pipe. The camera is positioned on one side of the pipe; as the fish swims from one connecting vessel to another, an image of one side's fish-eye is captured. When the fish swims back or a camera is positioned on the other side, an image of the other side's fish-eye is captured. Subsequently, iris localization and segmentation are performed, normalization is applied, and log-Gabor filters are used to extract iris texture features. KNN (Known Nearest Neighbor) algorithm, support vector machine, and Siamese network are used to classify the feature data within a small sample size. An adaptive weighted voting fusion of these three identification methods is then applied to achieve accurate identification of individual fish. Attached Figure Description

[0026] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0027] Figure 1 This is a diagram of a non-contact in-situ fish iris collection device according to Embodiment 1 of the present invention;

[0028] Figure 2 This is a flowchart of a non-contact in-situ fish iris collection device and individual identification method according to Embodiment 1 of the present invention;

[0029] Figure 3 This is a schematic diagram illustrating the annotation of the training set images for the fish individual iris segmentation module according to Embodiment 1 of the present invention;

[0030] Figure 4 This is a schematic diagram of the iris mask of an individual fish according to Embodiment 1 of the present invention;

[0031] Figure 5 This is a normalized schematic diagram of Embodiment 1 of the present invention;

[0032] Figure 6 This is a schematic diagram of the normalized iris image according to Embodiment 1 of the present invention;

[0033] Figures 7(a)-7(c)This is a schematic diagram of the iris image after texture feature extraction in Example 1. Detailed Implementation

[0034] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0035] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments of the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. Furthermore, it should be understood that the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0036] Example 1

[0037] This embodiment provides a non-contact fish individual identification method based on iris images;

[0038] like Figure 2 As shown, a non-contact fish individual identification method based on iris images includes:

[0039] S101: Acquire an image of the fish to be identified;

[0040] S102: Perform iris segmentation on the image of the fish to be identified to obtain an iris image;

[0041] S103: Normalize the iris image and extract texture features from the normalized iris image;

[0042] S104: Input the texture features into three classifiers respectively, and perform adaptive weighted voting fusion on the recognition results of the three classifiers to obtain the recognition result of the individual fish: when the recognition results of the three classifiers are consistent, directly output the recognition result of the individual fish; when the recognition results of the three classifiers are inconsistent, multiply the recognition probability value of each classifier by the corresponding classifier weight coefficient to obtain the final probability value of each classifier, and select the recognition result of the classifier with the largest final probability value as the final recognition result output.

[0043] Furthermore, such as Figure 1As shown, step S101: acquiring an image of the fish to be identified is performed using an image acquisition device. The image acquisition device includes a first communicating vessel and a second communicating vessel, which are connected by a connecting pipe. The first and second communicating vessels are used to place several fish to be detected. The fish can swim back and forth within the first communicating vessel, the connecting pipe, and the second communicating vessel. The connecting pipe is a square transparent pipe, and the diagonal length of the cross-section of the connecting pipe is a set value. When the fish swims through the connecting pipe, a picture is taken of the side of the fish.

[0044] Industrial cameras are installed on both the left and right sides of the connecting pipe. The industrial cameras use telecentric lenses and are connected to a power source. The industrial cameras upload the captured fish images to a host computer, which then identifies the individual fish.

[0045] It should be understood that the image acquisition device includes a connecting pipe (the connecting pipe is a square transparent pipe) connecting two communicating vessels, an industrial camera, a telecentric optical lens, a power supply, a host computer, and an LED light source. The camera is fixed to one side of the connecting pipe and uses a telecentric optical lens. The LED light source is fixed between the telecentric optical lens and the connecting pipe. The telecentric optical path design of the telecentric lens ensures that no parallax occurs during imaging. In addition, the telecentric lens has a large depth of field, and products within the field of view can obtain clear images regardless of adjustment, effectively avoiding interference from refraction and reflection. When a fish swims through the connecting pipe, the camera acquires an image, transmits it to the host computer, and stores it.

[0046] The most common method for individual fish identification currently relies on tools to mark and measure fish body parameters. This method is not only time-consuming and labor-intensive, but also causes harm to the fish. Therefore, this application provides a non-contact in-situ fish iris acquisition device and individual identification method. An LED light source and a camera are placed sequentially on one side of a connecting pipe connecting two connectors. When a fish swims through the connecting pipe, the image information of the fish's iris in the underwater scene is acquired, uploaded to a host computer, and stored, avoiding contact with the individual fish and preventing interference with their normal life.

[0047] Further, S102: Perform iris segmentation on the image of the fish to be identified to obtain an iris image, specifically using the trained U... 2 The Net network model performs iris segmentation on the image of the fish to be identified, and obtains the iris image.

[0048] Furthermore, the trained U 2 The training process for the Net network model includes:

[0049] Construct a training set, which consists of fish body images with known iris images;

[0050] Input the training set into U 2 In the Net network model, for U 2 The Net network model is trained, and training stops when the number of iterations exceeds a set number or the model's loss function value no longer decreases, yielding the trained U. 2 Net network model.

[0051] It should be understood that the construction of the training set involves using LabelMe software to label the iris portion, generating an iris mask, and creating an iris segmentation dataset. For the fish individuals in the training set images, ring annotations are performed on the irises using the LabelMe annotation tool, with the method "create circle" and the label "iris". Figure 3 The annotation is performed as shown. After annotation, a JSON-formatted annotation file is generated. This file is then preprocessed to generate an iris mask, which is input into the individual fish iris segmentation model for training. The iris mask is shown below. Figure 4 As shown.

[0052] It should be understood that the fish's eye occupies a very small area in the entire image, and due to differences in acquisition equipment, environment, and conditions, there will be different spectra and various types of noise interference, such as motion blur, specular reflection, and similar iris and pupil colors. This leads to inaccurate localization, lack of universality, and high computational cost and time consumption. The acquired images are compiled into an iris segmentation dataset, and a segmentation network is used for iris localization and segmentation.

[0053] Furthermore, S103: Normalizing the iris image refers to:

[0054] A coordinate transformation is performed on the annular iris region, mapping the iris region from the polar coordinate system to the rectangular coordinate system.

[0055] It should be understood that iris normalization refers to transforming the annular iris region by mapping it from polar coordinates to a rectangular iris region of size M×N in a Cartesian coordinate system. The mapping transformation process can be described as follows:

[0056] I(r,θ)→I(x(r,θ),y(t,θ)),

[0057] Where I(r,θ) is a polar coordinate system with the pupil center point as the origin, and I(x,y) is a normalized rectangular coordinate system. The iris normalization process is as follows: Figure 5 As shown, the normalized iris image is as follows: Figure 6 As shown.

[0058] Further, S103: Extract texture features from the normalized iris image, including:

[0059] The iris image in Cartesian coordinates is input into a multi-channel Gabor filter to obtain the results of the odd Gabor filter and the even Gabor filter for each pixel.

[0060] The modulus of the odd and even filters for each pixel is calculated based on the results of the odd Gabor filter and the even Gabor filter for each pixel.

[0061] Calculate the mean and variance of the moduli of the odd and even filters for all pixels, and use the mean and variance as texture features.

[0062] A multi-channel Gabor filter decomposes the image to be processed into a set of filtered global images that reflect texture features according to different frequencies and directions. The filter parameters of the multi-channel Gabor filter are wavelength = 1 / f0 = 3. When θ = 45°, 90°, and 135°, the modulus of the odd and even filters is taken, and their mean and variance are calculated as the feature data for classification. Texture features, such as… Figures 7(a)-7(c) As shown, the modulus of the odd / even filter is taken, and its mean and variance are calculated. Here, wavelength represents the wavelength in pixels, f0 represents the frequency, and θ represents the tilt angle of the Gabor function graph in degrees.

[0063] Two-dimensional Gabor function:

[0064]

[0065] Where x and y represent pixel coordinates, λ represents wavelength, θ represents the tilt angle of the Gabor function graph, φ represents phase offset, σ represents the standard deviation of the Gaussian function, γ represents aspect ratio, and σ x =kx*wavelength, σ y =ky*wavelength, kx, ky are the scaling factors of the filter relative to the filter wavelength, x' = x cosθ + y sinθ, y' = -x sinθ + y cosθ.

[0066] Further, the step of calculating the modulus of the odd / even filter for each pixel based on the results of the odd Gabor filter and the even Gabor filter for each pixel specifically includes:

[0067] Results of the odd Gabor filter:

[0068]

[0069] Where x and y represent pixel coordinates, λ represents wavelength, φ represents phase offset, and σx =kx*wavelength, σ y =ky*wavelength, kx,ky are the scaling factors of the filter relative to the filter wavelength, x' = xcosθ + ysinθ, y' = -xsinθ + ycosθ; θ represents the tilt angle of the Gabor function graph;

[0070] Results of even Gabor filter filtering:

[0071]

[0072] The modulus Aim of the parity filter:

[0073]

[0074] Further, in S104: the texture features are input into three classifiers respectively, and the recognition results of the three classifiers are adaptively weighted and fused to obtain the recognition result of the individual fish: when the recognition results of the three classifiers are consistent, the recognition result of the individual fish is directly output; when the recognition results of the three classifiers are inconsistent, the final probability value of each classifier is obtained by multiplying the recognition probability value of each classifier by the weight coefficient of the corresponding classifier, and the recognition result of the classifier with the largest final probability value is selected as the final recognition result output. The three classifiers are KNN, support vector machine and Siamese network respectively.

[0075] Furthermore, the weight coefficients w of the classifier i The calculation process includes:

[0076]

[0077]

[0078] Where, p ij H represents the classification accuracy of the i-th classifier for the known j-th fish. i (x) represents the Shannon entropy value, and n represents the total number of fish.

[0079] Given the classification accuracy of the j-th fish, which is a historical time point, and the prediction accuracy of the classifier for the fish given the fish number, we have the following information.

[0080] It should be understood that the weight coefficients of the classifiers are calculated by taking the fish identification accuracy of each classifier at a historical time point with the fish number known, and then calculating the weight coefficients of each classifier based on the fish identification accuracy of each classifier.

[0081] It should be understood that the embodiments of the present invention can monitor the growth of each fish (numbered 001, 002, ... 003) in the same batch of fish in a fishpond at different time points (January, February, March... December).

[0082] Example 2

[0083] This embodiment provides a non-contact fish individual identification system based on iris images;

[0084] A non-contact fish identification system based on iris images includes:

[0085] The acquisition module is configured to acquire an image of the fish to be identified.

[0086] The iris segmentation module is configured to perform iris segmentation on the image of the fish to be identified, thereby obtaining an iris image;

[0087] The feature extraction module is configured to: normalize the iris image and extract texture features from the normalized iris image;

[0088] The weighted fusion module is configured to: input texture features into three classifiers respectively, perform adaptive weighted voting fusion on the recognition results of the three classifiers to obtain the recognition result of the individual fish; when the recognition results of the three classifiers are consistent, the recognition result of the individual fish is directly output; when the recognition results of the three classifiers are inconsistent, the final probability value of each classifier is obtained by multiplying the recognition probability value of each classifier by the corresponding weight coefficient, and the recognition result of the classifier with the largest final probability value is selected as the final recognition result output.

[0089] It should be noted that the acquisition module, iris segmentation module, feature extraction module, and weighted fusion module described above correspond to steps S101 to S104 in Embodiment 1. The examples and application scenarios implemented by these modules and their corresponding steps are the same, but they are not limited to the content disclosed in Embodiment 1. It should also be noted that these modules, as part of the system, can be executed in a computer system, such as a set of computer-executable instructions.

[0090] The descriptions of each embodiment in the above embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0091] The proposed system can be implemented in other ways. For example, the system embodiments described above are merely illustrative, and the division of modules described above is only a logical functional division. In actual implementation, there may be other division methods. For example, multiple modules may be combined or integrated into another system, or some features may be ignored or not executed.

[0092] Example 3

[0093] This embodiment also provides an electronic device, including: one or more processors, one or more memories, and one or more computer programs; wherein, the processor is connected to the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory to cause the electronic device to perform the method described in Embodiment 1.

[0094] It should be understood that in this embodiment, the processor can be a central processing unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor, etc.

[0095] The memory may include read-only memory and random access memory, providing instructions and data to the processor. A portion of the memory may also include non-volatile random access memory. For example, the memory may also store device type information. In implementation, the steps of the above method can be performed by integrated logic circuits in the processor's hardware or by instructions in software form.

[0096] The method in Embodiment 1 can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor. The software modules can reside in readily available storage media in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory; the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method. To avoid repetition, a detailed description is not provided here.

[0097] Those skilled in the art will recognize that the units and algorithm steps described in connection with the various examples of this embodiment can be implemented in electronic hardware or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0098] Example 4 This example also provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, complete the method described in Example 1.

[0099] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A non-contact fish individual identification method based on an iris image, characterized by, The method comprises the following steps: obtaining an image of a fish to be identified; performing iris segmentation on the image of the fish to be identified to obtain an iris image; performing normalization processing on the iris image, and performing texture feature extraction on the normalized iris image, which comprises the following steps: inputting the iris image in the rectangular coordinate system into a multi-channel filter Gabor to obtain the filtering result of each pixel point filtered by the odd Gabor filter and the filtering result of each pixel point filtered by the even Gabor filter; calculating the modulus of the odd-even filter of each pixel point according to the filtering result of each pixel point filtered by the odd Gabor filter and the filtering result of each pixel point filtered by the even Gabor filter, which comprises the following steps: the filtering result of the odd Gabor filter: ； wherein, , respectively represent pixel coordinate positions, represents a wavelength, represents a phase shift amount, , , is a scaling factor of the filter with respect to the filter wavelength, ; denotes the tilt angle of the Gabor and function images; the filtering result of the even Gabor filter: ； odd-even filter module : ； calculating the mean value and the variance of the modulus of the odd-even filter of all pixel points, and taking the mean value and the variance as the texture feature; inputting the texture feature into three classifiers respectively, performing adaptive weighted voting fusion on the identification results of the three classifiers, and obtaining the identification result of the fish individual: when the identification results of the three classifiers are consistent, directly outputting the identification result of the fish individual; when the identification results of the three classifiers are inconsistent, multiplying the identification probability value of each classifier by the weight coefficient of the corresponding classifier to obtain the final probability value of each classifier, selecting the identification result of the classifier with the maximum final probability value as the final identification result, and outputting the final identification result.

2. The non-contact fish individual identification method based on an iris image according to claim 1, wherein The image of the fish to be identified is obtained by an image acquisition device, the image acquisition device comprises a first communicating vessel and a second communicating vessel, the first communicating vessel and the second communicating vessel are connected through a communicating pipeline, the first communicating vessel and the second communicating vessel are used for placing a plurality of detected fish, and the fish can move back and forth in the first communicating vessel, the communicating pipeline and the second communicating vessel; the communicating pipeline is a square transparent pipeline, the length of the diagonal of the cross section of the communicating pipeline is a set value; the side of the fish is photographed during the process of the fish swimming through the communicating pipeline; industrial cameras are installed on the left and right sides of the communicating pipeline, the industrial cameras adopt optical telecentric lenses, the industrial cameras are connected with a power supply, and the industrial cameras upload the collected fish image to an upper computer, and the upper computer identifies the individual fish.

3. The non-contact fish individual identification method based on an iris image according to Claim 1, wherein For the image of the fish to be identified, iris segmentation is performed to obtain an iris image, and the trained U ２ Net network model is used to segment the iris of the image of the fish to be identified to obtain an iris image. The normalization processing on the iris image refers to coordinate transformation on the annular iris region, and mapping the iris region from the polar coordinate system to the rectangular coordinate system.

4. The non-contact fish individual identification method based on an iris image according to claim 3, characterized by, The trained U ２ The Net network model, the training process comprises: A training set is constructed, and the training set is an image of a fish body of a known fish iris image; The training set is input to the U ２ In the U ２ Net network model is trained, and when the number of iterations exceeds a set number of times or the loss function value of the model no longer decreases, the training is stopped, and the trained U ２ Net network model.

5. The non-contact fish individual identification method based on an iris image according to Claim 1, wherein weight coefficients of the classifier the computing process comprises: wherein, represents the classification accuracy of the i-th classifier for the known j-th fish, represents the Shannon entropy value, represents the total number of fish.

6. A non-contact fish individual identification system based on an iris image, characterized by, The method comprises the following steps: an acquisition module configured to obtain an image of a fish to be identified; an iris segmentation module configured to perform iris segmentation on the image of the fish to be identified to obtain an iris image; a feature extraction module configured to perform normalization processing on the iris image, and perform texture feature extraction on the normalized iris image, which comprises the following steps: inputting the iris image in the rectangular coordinate system into a multi-channel filter Gabor to obtain the filtering result of each pixel point filtered by the odd Gabor filter and the filtering result of each pixel point filtered by the even Gabor filter; According to the result of the odd Gabor filter filtering of each pixel point and the result of the even Gabor filter filtering of each pixel point, the modulus of the odd-even filter of each pixel point is calculated, and the modulus of the odd-even filter of each pixel point is calculated. The result of the odd Gabor filter filtering is: ； wherein , respectively represent pixel coordinate positions, represents a wavelength, represents a phase shift amount, , , is a scale factor of the filter with respect to the filter wavelength, ; denotes the tilt angle of the Gabor and function images; The result of the even Gabor filter filtering is: ； odd filter : ； The mean and variance of the modulus of the odd-even filter of all pixel points are calculated, and the mean and variance are taken as the texture feature. The weighted fusion module is configured to input the texture feature into three classifiers respectively, to perform adaptive weighted voting fusion on the identification results of the three classifiers, and to obtain the identification result of the fish individual: when the identification results of the three classifiers are consistent, the identification result of the fish individual is directly output; when the identification results of the three classifiers are inconsistent, the identification probability value of each classifier is multiplied by the weight coefficient of the corresponding classifier to obtain the final probability value of each classifier, and the identification result of the classifier with the maximum final probability value is selected as the final identification result.

7. An electronic device, comprising: a memory for non-transiently storing computer readable instructions; and a processor for running the computer readable instructions, wherein the computer readable instructions, when executed by the processor, perform the method of any one of claims 1-5.

8. A storage medium characterized by, non-transiently storing computer readable instructions, wherein when the non-transient computer readable instructions are executed by a computer, the instructions perform the method of any one of claims 1-5.

Images