Iris image segmentation method and device, equipment and storage medium

Abstract

The invention provides an iris image segmentation method, and relates to the technical field of iris recognition, and the method comprises the steps: obtaining a plurality of iris images, carrying out the data preprocessing of the plurality of iris images, and dividing the preprocessed iris images into a training data set and a verification data set; constructing an iris image segmentation deep neural network model, training the iris image segmentation deep neural network model according to the training data set, and verifying the iris image segmentation deep neural network model through the verification data set to obtain a trained iris image segmentation deep neural network model; inputting an iris image to be segmented into the trained iris image segmentation deep neural network model for forward reasoning to obtain a current reasoning result; and post-processing the current inference result, and respectively extracting binary mask images of the background, the sclera, the iris and the pupil. In this way, iris segmentation with higher precision is achieved, and the problem of'exceptional blocks' is effectively solved.

Description

Technical Field

[0001] The present application relates to the technical field of iris recognition, and in particular to an iris image segmentation method, device, equipment and storage medium. Background Art

[0002] Iris recognition technology is a technology that uses the iris in the human eye to authenticate identity. It is a type of human biometric technology. Iris recognition has the characteristics of uniqueness, stability, non-contact, and high security. It is recognized as the most accurate and convenient biometric technology and has been widely used in finance, security, checkpoints, access control, insurance and other scenarios that require accurate identity authentication.

[0003] Iris segmentation is a processing technology that removes noise from iris images and extracts effective iris areas. It is one of the key technologies to ensure the accuracy of iris recognition. Traditional iris segmentation methods rely on prior knowledge, have high computational complexity, and weak noise resistance. They have gradually been replaced by iris segmentation methods based on deep learning. However, even iris segmentation methods based on deep learning still have some obvious problems. The first is the insufficient segmentation accuracy. The existing models generally use CE Loss and multi-classification methods to obtain the binary masks of the iris and pupil, while the accuracy is evaluated by IOU. There is no direct correlation between the two. The second is the "exceptional block" problem. For example, some blocks in the non-iris area are mistakenly segmented as iris areas. Although the method of using the spatial inclusion relationship between the iris and the pupil can solve this problem, there are still theoretical deficiencies. For example, when the pupil is partially occluded, this spatial inclusion relationship does not exist. Summary of the invention

[0004] The present application provides an iris image segmentation method, apparatus, device and storage medium, which achieves higher-precision iris segmentation and effectively solves the "exceptional block" problem by reasonably using a combination of multiple loss functions.

[0005] According to a first aspect of the present application, a method for iris image segmentation is provided, the method comprising: Acquire multiple iris images, perform data preprocessing on the multiple iris images, and divide the preprocessed iris images into a training data set and a verification data set; Constructing a deep neural network model for iris image segmentation, and training the deep neural network model for iris image segmentation according to the training data set, and verifying the model through the verification data set to obtain a trained deep neural network model for iris image segmentation; Inputting the iris image to be segmented into the trained iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result; The current inference result is post-processed to extract binary mask images of the background, sclera, iris and pupil respectively.

[0006] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the data preprocessing of the multiple iris images comprises: Dividing each iris image into a plurality of different and mutually independent regions, and marking each pixel point in each iris image according to each region to obtain each marked iris image; Performing data augmentation processing on each of the annotated iris images; The image after data augmentation processing is scaled to a single-channel grayscale image of a preset size, and the single-channel grayscale image is standardized to obtain a preprocessed iris image.

[0007] According to the above aspects and any possible implementation, an implementation is further provided, wherein the step of constructing a deep neural network model for iris image segmentation includes: A symmetrical network structure is used to design a deep neural network model structure for iris image segmentation, wherein the symmetrical network structure includes an encoder and a decoder; Determining the training loss of the iris image segmentation deep neural network model structure by combining multiple losses with different weights; According to the iris image segmentation deep neural network model structure and the iris image segmentation deep neural network model structure training loss, an iris image segmentation deep neural network model is constructed.

[0008] According to the above aspects and any possible implementation, an implementation is further provided, wherein the loss of the iris image segmentation deep neural network model structure training is determined by combining multiple losses with different weights, including: Determine a joint loss function according to the loss values ​​calculated by the CE Loss function, the Dice Loss function and the Hausdorff loss function and their corresponding loss calculation weights, and use the joint loss function as the training loss of the iris image segmentation deep neural network model structure; Among them, the CE Loss function is used to calculate the difference loss value between the probability distribution of the model prediction output and the true category label; the Dice Loss function is used to calculate the loss value between the predicted output area and the labeled area; the Hausdorffloss function calculates the distance loss value between the predicted output area and the labeled area.

[0009] According to the above aspects and any possible implementation, an implementation is further provided, wherein the joint loss function is: ; in, represents the total joint loss value, Represents each loss item, For the The calculation weight of the loss, For the The calculated value of the loss.

[0010] According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the training of the iris image segmentation deep neural network model according to the training data set comprises: According to the above aspects and any possible implementation manner, an implementation manner is further provided, wherein the training of the iris image segmentation deep neural network model according to the training data set comprises: Step S61: input the training data set into the iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result; Step S62: According to the current reasoning result and the real annotation information, the loss is calculated by the joint loss function to obtain the current joint loss; Step S63: The current joint loss is transmitted back to the iris image segmentation deep neural network model through the reverse gradient propagation mechanism, and the current network weight is adjusted in combination with the learning rate; The steps S61 to S63 are executed cyclically until the training reaches a specified preset number of cycles, and the training ends.

[0011] According to the above aspects and any possible implementation, an implementation is further provided, wherein the inputting of the training data set into the iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result comprises: The encoder extracts features from the input training data set and maps it to the latent space to obtain latent space data; The latent space data is recovered by a decoder, and detail features are obtained by peer layer jumps, and finally multi-channel image data of the same size as the input image is output; The output image is processed by the softmax function to obtain the probability prediction results of multiple classifications.

[0012] According to a second aspect of the present application, an iris image segmentation device is provided, comprising: A data preprocessing module, used for acquiring a plurality of iris images, performing data preprocessing on the plurality of iris images, and dividing the preprocessed iris images into a training data set and a verification data set; A neural network model construction module is used to construct a deep neural network model for iris image segmentation, and train the deep neural network model for iris image segmentation according to the training data set, and verify it through the verification data set to obtain a trained deep neural network model for iris image segmentation; A forward reasoning module, used for inputting the iris image to be segmented into the trained iris image segmentation deep neural network model to perform forward reasoning and obtain the current reasoning result; The binary mask image extraction module is used to post-process the current reasoning result and extract the binary mask images of the background, sclera, iris and pupil respectively.

[0013] According to a third aspect of the present application, an electronic device is provided, comprising: a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the program, the method described above is implemented.

[0014] According to a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the method according to the first aspect of the present application is implemented.

[0015] Compared with the prior art, the present invention achieves the following beneficial effects: (1) During the data labeling stage, the iris image is labeled with four regions: background, sclera, iris, and pupil. The number of labeled images should be sufficient and diverse, covering different populations, collection environments, facilities, and occlusion conditions, so that the model can adapt to a variety of actual scenarios, enhance generalization capabilities, and reduce performance degradation caused by data differences.

[0016] (2) Improve segmentation accuracy through joint loss function design. CE Loss is used to ensure correct classification results and enhance robustness; Dice Loss constrains the classification range, which leads to more accurate segmentation because it is sensitive to segmentation errors; Hausdorff loss constrains the distance of segmentation results, prompting the model to generate high-precision segmentation results, effectively overcoming the problem of insufficient accuracy of traditional methods and some deep learning methods.

[0017] It should be understood that the contents described in the Summary of the Invention are not intended to limit the key or important features of the embodiments of the present application, nor are they intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The above and other features, advantages and aspects of the embodiments of the present application will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present application. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, among which: Figure 1 A flow chart of an iris image segmentation method according to an embodiment of the present application is shown; Figure 2 An example diagram of iris image segmentation marking according to an embodiment of the present application is shown; Figure 3 A structural diagram of a deep neural network model according to an embodiment of the present application is shown; Figure 4 A ConvBlock structure diagram according to an embodiment of the present application is shown; Figure 5 A DownBlock structure diagram according to an embodiment of the present application is shown; Figure 6 shows a structural diagram of UpBlock according to an embodiment of the present application; Figure 7 shows a BaseBlock structure diagram according to an embodiment of the present application; Figure 8 An example diagram of the loss combination effect according to an embodiment of the present application is shown; Fig. 9 A block diagram of an iris image segmentation device according to an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0020] In addition, the term "and / or" in this article is only a description of the association relationship between the associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0021] Figure 1 A flowchart of a method 100 for processing an iris image segmentation is shown. Figure 1 As shown, the method comprises the following steps: S110, acquiring a plurality of iris images, performing data preprocessing on the plurality of iris images, and dividing the preprocessed iris images into a training data set and a verification data set.

[0022] In some embodiments, in order to ensure that the model can adapt to a variety of actual scenarios and enhance generalization capabilities, the number of annotated images should be sufficient and diverse. Therefore, a sufficient number of iris images should be obtained, and the iris images should have diverse characteristics. The diverse characteristics need to cover different populations, different collection environments, different collection facilities and different occlusion conditions including but not limited to eyelids, eyelashes, hair, external reflections, different wearing items (such as frame glasses, contact lenses, etc.), etc.

[0023] In some embodiments, performing data preprocessing on a plurality of iris images includes: (1) Data annotation is performed on multiple iris images. Specifically, each iris image is divided into multiple different and independent regions, for example, into four regions: background, sclera, iris, and pupil. Each pixel in each iris image is annotated according to each divided region (background, sclera, iris, and pupil) to ensure that any point (pix) in the image belongs to and can only belong to one of the above four regions. The labeled result image is as follows: Figure 2 As shown in the figure, the labeled result image uses one-shot encoding, and the corresponding values ​​of the background, sclera, iris, and pupil are 0, 1, 2, and 3 respectively.

[0024] (2) performing data augmentation processing on the annotated iris images, such as horizontal flipping, rotation, etc. It should be emphasized that if the data augmentation processing affects the annotated information of the pixels, the corresponding annotated information should be adjusted at the same time.

[0025] (3) Scaling the image after data augmentation processing into a single-channel grayscale image of a preset size, and standardizing the single-channel grayscale image, the standardization processing method is: ; in, is the pixel value of the image after normalization. is the image pixel value before normalization, is the statistical average of the pixel values ​​of the training data set, is the standard deviation of the pixel values ​​in the training dataset.

[0026] In some embodiments, the preprocessed iris image may be divided into a training data set and a verification data set in a certain ratio, for example, the ratio of the training data set to the verification data set is 9:1.

[0027] S120, constructing a deep neural network model for iris image segmentation, and training the deep neural network model for iris image segmentation according to the training data set, and verifying the model through the verification data set to obtain a trained deep neural network model for iris image segmentation.

[0028] In some embodiments, constructing the iris image segmentation deep neural network model includes designing the iris image segmentation deep neural network model network structure and the iris image segmentation deep neural network model structure training loss. Specifically, designing the iris image segmentation deep neural network model network structure includes the following steps: (1) The deep neural network model for iris image segmentation adopts a symmetrical network structure, which is divided into two parts: an encoder and a decoder. The model structure in this embodiment is as follows: Figure 3 The detailed information of the model structure is shown in the Appendix. Figure 4 -Attached Figure 7 As shown, the following Table 1 shows the detailed network structure: Table 1: ; (2) The encoder is responsible for feature extraction of the input data, including multiple downsampling modules, mapping the input data to the corresponding latent space; The decoder is responsible for recovering the latent space data, including multiple upsampling modules, and obtains detailed features with the help of peer layer jumpers, and finally obtains 4-channel image data of the same size as the input image.

[0029] (3) The output image is processed by the softmax function to obtain the probability prediction results corresponding to the four categories.

[0030] The design of the training loss of the deep neural network model structure for iris image segmentation includes: determining a joint loss function according to the loss values ​​calculated by the CE Loss function, the DiceLoss function and the Hausdorff loss function and their corresponding loss calculation weights, and using the joint loss function as the training loss of the deep neural network model structure for iris image segmentation; Among them, the CE Loss function is used to calculate the difference loss value between the probability distribution of the model prediction output and the true category label , the formula is: ; in, is the category label. If the current category is ,but , otherwise it is equal to 0; is the output of the deep neural network model, which means the predicted category is The output value is calculated using the softmax function mentioned above.

[0031] The Dice Loss function is used to calculate the loss value between the predicted output area and the marked area , the formula is: ; in, , Respectively represent a set of pixels of a certain type of prediction result; express and The intersection of . , Separately With Collection The total number of pixels.

[0032] The Hausdorff loss function is used to calculate the distance loss value between the predicted output area and the marked area , the formula is: ; ; ; in, For two sets, Representing a collection A point in Representing a collection A point in For point , The Euclidean distance between is the maximum distance from each point in set A to the farthest point in set B, is the maximum distance from each point in set B to the farthest point in set A, is the distance between set A and set B, take and The maximum value of .

[0033] The loss values ​​calculated by the above CE Loss function, Dice Loss function and Hausdorff loss function are superimposed according to different weights, and the calculation method is: ; in, represents the total joint loss value, Represents each loss item, with values ​​of 1, 2, and 3. For the The calculation weight of the loss, For the The calculated value of the loss term.

[0034] CE loss tends to provide correct classification results and increase the distance between the positive and negative prediction results, which is conducive to improving the robustness of the segmentation results. However, it is not sensitive to the range of misclassification. Dice loss can be seen as a variant of IOU loss, which tends to constrain the classification range. And compared with IOU loss, Dice loss is more sensitive to segmentation errors, and thus tends to guide the model to produce more accurate segmentation results. However, it is not sensitive to the distance of misclassification. Hausdorff loss imposes distance constraints on the segmentation results, is sensitive to "exceptional blocks", and can greatly reduce the possibility of "exceptional blocks". Therefore, through the combination of the above functions, the model can be prompted to generate high-precision segmentation results, effectively overcoming the problem of insufficient accuracy of traditional methods and some deep learning methods.

[0035] In some embodiments, training the iris image segmentation deep neural network model according to the training data set comprises the following steps: Input the training data set into the iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result; According to the current inference result and the real annotation information, the loss is calculated by the joint loss function to obtain the current joint loss; The current joint loss is propagated back to the iris image segmentation deep neural network model through the reverse gradient propagation mechanism, and the current network weight is adjusted in combination with the learning rate; The above steps are executed repeatedly until the training reaches the specified preset number of cycles, and the training is completed to obtain a trained iris image segmentation deep neural network model.

[0036] At the same time, during the training process, the validation data set is used to verify the network model and screen parameters. The network model parameter screening conditions are: the indicators of the validation data set are the best results of the current training and meet expectations.

[0037] S130, inputting the iris image to be segmented into the trained iris image segmentation deep neural network model to perform forward reasoning to obtain a current reasoning result.

[0038] In some embodiments, the iris image to be segmented is first preprocessed in the same manner as that used in step S110 above. The preprocessed iris image is input into the trained iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result.

[0039] S140, post-processing the current inference result to extract binary mask images of the background, sclera, iris and pupil respectively.

[0040] In some embodiments, the category to which each pixel belongs may be determined according to the current reasoning result, for example, determining that the category to which each pixel belongs is background, sclera, iris, or pupil; Convert pixels belonging to different categories into binary form. For example, pixels belonging to the background may be assigned a value of 0; pixels belonging to the sclera may be assigned a value of 1; pixels belonging to the iris may be assigned a value of 2; and pixels belonging to the pupil may be assigned a value of 3 (refer to the settings in the data annotation stage), thereby generating a preliminary binary image; In addition, in order to further optimize the mask image, an erosion operation can be performed to remove small noise points or isolated pixels, and a dilation operation can be performed to fill small holes or connect adjacent areas to make the contours of each area clearer and more continuous, thereby obtaining a more accurate binary mask image of the background, sclera, iris and pupil.

[0041] It should be emphasized that in this embodiment, according to the uniqueness principle of pixel classification, each pixel takes the classification result with the largest different classification prediction results as the final prediction result. The final prediction result is exemplified by Figure 8 As shown in the figure, for the case without light pollution, the use of the joint loss function can significantly eliminate the "exceptional blocks", such as the first two pictures; for the case with severe light pollution, not only can the "exceptional blocks" be eliminated, but the segmentation accuracy is also significantly improved, such as the last three pictures.

[0042] According to the embodiments of the present disclosure, the following beneficial effects are achieved: (1) During the data labeling stage, the iris image is labeled with four regions: background, sclera, iris, and pupil. The number of labeled images should be sufficient and diverse, covering different populations, collection environments, facilities, and occlusion conditions, so that the model can adapt to a variety of actual scenarios, enhance generalization capabilities, and reduce performance degradation caused by data differences.

[0043] (2) Improve segmentation accuracy through joint loss function design. CE Loss is used to ensure correct classification results and enhance robustness; Dice Loss constrains the classification range, which leads to more accurate segmentation because it is sensitive to segmentation errors; Hausdorff loss constrains the distance of segmentation results, prompting the model to generate high-precision segmentation results, effectively overcoming the problem of insufficient accuracy of traditional methods and some deep learning methods.

[0044] (3) By utilizing the characteristic of Hausdorff loss that is sensitive to “exceptional blocks”, the probability of non-iris areas being incorrectly segmented as iris areas is greatly reduced, thereby improving the defects of the existing model in this regard. Even in complex situations such as partial occlusion of the pupil, incorrect segmentation can be effectively avoided.

[0045] It should be noted that, for the aforementioned method embodiments, for the sake of simplicity, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the present application is not limited by the described order of actions, because according to the present application, certain steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily required by the present application.

[0046] The above is an introduction to the method embodiment. The following is a further explanation of the scheme described in this application through an apparatus embodiment.

[0047] Fig. 9 FIG. 9 is a block diagram of an iris image segmentation device 900 according to an embodiment of the present application. Fig. 9 As shown, the device 900 includes: A data preprocessing module 910 is used to obtain a plurality of iris images, perform data preprocessing on the plurality of iris images, and divide the preprocessed iris images into a training data set and a verification data set; A neural network model construction module 920 is used to construct a deep neural network model for iris image segmentation, and train the deep neural network model for iris image segmentation according to the training data set, and verify it through the verification data set to obtain a trained deep neural network model for iris image segmentation; A forward reasoning module 930 is used to input the iris image to be segmented into the trained iris image segmentation deep neural network model to perform forward reasoning to obtain a current reasoning result; The binary mask image extraction module 940 is used to post-process the current reasoning result and extract the binary mask images of the background, sclera, iris and pupil respectively.

[0048] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the described module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0049] In the technical solution of this application, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0050] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution disclosed in this application can be achieved, and this document is not limited here.

[0051] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A method for segmenting an iris image, characterized in that: include: Acquire multiple iris images, perform data preprocessing on the multiple iris images, and divide the preprocessed iris images into a training data set and a verification data set; Constructing a deep neural network model for iris image segmentation, and training the deep neural network model for iris image segmentation according to the training data set, and verifying the model through the verification data set to obtain a trained deep neural network model for iris image segmentation; Inputting the iris image to be segmented into the trained iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result; The current inference result is post-processed to extract binary mask images of the background, sclera, iris and pupil respectively.

2. The method according to claim 1, characterized in that: The data preprocessing of the multiple iris images includes: Dividing each iris image into a plurality of different and mutually independent regions, and marking each pixel point in each iris image according to each region to obtain each marked iris image; Performing data augmentation processing on each of the annotated iris images; The image after data augmentation processing is scaled to a single-channel grayscale image of a preset size, and the single-channel grayscale image is standardized to obtain a preprocessed iris image.

3. The method according to claim 1, characterized in that The method of constructing a deep neural network model for iris image segmentation includes: A symmetrical network structure is used to design a deep neural network model structure for iris image segmentation, wherein the symmetrical network structure includes an encoder and a decoder; Determining the training loss of the iris image segmentation deep neural network model structure by combining multiple losses with different weights; According to the iris image segmentation deep neural network model structure and the iris image segmentation deep neural network model structure training loss, an iris image segmentation deep neural network model is constructed.

4. The method according to claim 3, characterized in that The method of determining the training loss of the iris image segmentation deep neural network model structure by combining multiple losses with different weights includes: Determine a joint loss function according to the loss values ​​calculated by the CE Loss function, the Dice Loss function and the Hausdorff loss function and their corresponding loss calculation weights, and use the joint loss function as the training loss of the iris image segmentation deep neural network model structure; Among them, the CE Loss function is used to calculate the difference loss value between the probability distribution of the model prediction output and the true category label; the Dice Loss function is used to calculate the loss value between the predicted output area and the labeled area; the Hausdorff loss function calculates the distance loss value between the predicted output area and the labeled area.

5. The method according to claim 4, characterized in that The joint loss function is: ； in, represents the total joint loss value, Represents each loss item, For the The loss calculation weight, For the The calculated value of the loss.

6. The method according to claim 4, characterized in that The step of training the iris image segmentation deep neural network model according to the training data set includes: Step S61: input the training data set into the iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result; Step S62: According to the current reasoning result and the real annotation information, the loss is calculated by the joint loss function to obtain the current joint loss; Step S63: The current joint loss is transmitted back to the iris image segmentation deep neural network model through the reverse gradient propagation mechanism, and the current network weight is adjusted in combination with the learning rate; The steps S61 to S63 are executed cyclically until the training reaches a specified preset number of cycles, and the training ends.

7. The method according to claim 6, characterized in that The step of inputting the training data set into the iris image segmentation deep neural network model for forward reasoning to obtain the current reasoning result includes: The encoder extracts features from the input training data set and maps it to the latent space to obtain latent space data; The latent space data is recovered by a decoder, and detail features are obtained by peer layer jumps, and finally multi-channel image data of the same size as the input image is output; The output image is processed by the softmax function to obtain the probability prediction results of multiple classifications.

8. An iris image segmentation device, characterized in that: include: A data preprocessing module, used for acquiring a plurality of iris images, performing data preprocessing on the plurality of iris images, and dividing the preprocessed iris images into a training data set and a verification data set; A neural network model construction module is used to construct a deep neural network model for iris image segmentation, and train the deep neural network model for iris image segmentation according to the training data set, and verify it through the verification data set to obtain a trained deep neural network model for iris image segmentation; A forward reasoning module, used for inputting the iris image to be segmented into the trained iris image segmentation deep neural network model to perform forward reasoning and obtain the current reasoning result; The binary mask image extraction module is used to post-process the current reasoning result and extract the binary mask images of the background, sclera, iris and pupil respectively.

9. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 7.

10. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-7.

Images