Iris image segmentation method and device, medium and equipment
Through deep learning model and multi-model integration technology, the problems of cumbersome and poor robustness of iris segmentation steps in the prior art are solved, and high-accurate iris segmentation under different lighting and angles are achieved.
Patent Information
- Application Number
- CN202310864994.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-07-13
- Publication Date
- 2025-01-14
AI Technical Summary
The prior art has complicated steps when acquiring the effective iris area, poor generalization and robustness, making it difficult to accurately segment under different light and angles.
The deep learning model is used to segment the iris image into effective areas that only include iris texture areas. Through the integration of multiple models and data augmentation technology, the accuracy and robustness of the segmentation are improved.
It realizes simple and clear iris segmentation, improves the accuracy and generalization of recognition, and can effectively deal with iris segmentation in complex scenarios such as strabismus and squinting.
Smart Images
Figure CN119313895A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of biometric identification, and in particular to an iris image segmentation method, device, medium and equipment. Background Art
[0002] As a widely used biometric recognition technology, iris recognition is accurate, stable, secure and reliable. A complete iris identity recognition system mainly includes iris image acquisition, iris image quality assessment, iris image preprocessing, iris image feature extraction and iris feature comparison. The purpose of iris image preprocessing is to obtain the effective area of iris for iris image feature extraction.
[0003] The method of obtaining the effective area of the iris in the prior art is to first locate the inner and outer circles of the iris, then perform iris noise detection, and finally segment the effective area of the iris. The inner circle of the iris is the boundary between the pupil and the iris, and the outer circle of the iris is the boundary between the iris and the sclera. Figure 1 This is a schematic diagram of the results of iris inner and outer circle positioning. Figure 1 It can be seen that the annular area of the iris contains a lot of noise, mainly eyelid cover, eyelash cover, light spot, etc. If these noise interferences are not eliminated and regarded as real iris information, it will inevitably cause deviations in the iris feature extraction results and lead to recognition errors. Iris noise detection mainly includes upper and lower eyelid positioning, light spot detection and eyelash detection, and finally generates a noise template.
[0004] After the iris inner and outer circle positioning, eyelid positioning, light spot detection and eyelash detection are completed, a noise template is generated to extract iris features. The so-called noise template refers to the results of eyelid positioning, eyelash detection and light spot detection represented by a binary [0,1] matrix of the same size as the iris image. The noise part of the iris is 0, and the iris and sclera parts are 1. Figure 2 This is the effect after positioning the inner and outer circles of the iris and the eyelids. Figure 3 This is the effect image of the generated noise template.
[0005] Through the positioning results of the inner and outer circles of the iris and the noise template, the part between the inner and outer circles of the iris and whose value is 1 on the noise template is selected to obtain the effective iris area.
[0006] The existing methods for obtaining the effective iris area (i.e., iris segmentation) have the disadvantages of cumbersome and complicated steps, poor generalization and robustness, etc. Summary of the invention
[0007] In order to solve the defects of the prior art, the present application provides an iris image segmentation method, device, medium and equipment. The segmentation method is simple and clear, and has high accuracy, generalization and robustness.
[0008] This application provides the following technical solutions:
[0009] In a first aspect, the present application provides an iris image segmentation method, the method comprising:
[0010] Obtain an iris image to be segmented;
[0011] The iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction that only includes the iris texture area is segmented from the iris image.
[0012] Furthermore, the iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction including only the iris texture area is segmented from the iris image, including:
[0013] Inputting the iris image into the deep learning model, encoding the iris image, and using deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map;
[0014] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area;
[0015] The value of each pixel point on the probability feature map is compared with the set segmentation threshold. If the value of the pixel point in the probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0016] Furthermore, there are multiple deep learning models, and each deep learning model is trained by a different training data set.
[0017] The iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction including only the iris texture area is segmented from the iris image, including:
[0018] Inputting the iris image into the multiple deep learning models respectively, encoding the iris image, and using deconvolution to expand the encoded feature map to the same size as the iris image, so as to obtain the same number of probability feature maps as the deep learning model;
[0019] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area;
[0020] Calculate the average value of all probability feature maps at the same position to obtain the average probability feature map;
[0021] The value of each pixel point on the average probability feature map is compared with the set segmentation threshold. If the value of the pixel point in the average probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0022] Furthermore, the calculation of the average value of all probability feature maps at the same position also includes:
[0023] Performing a transformation operation on the iris image to obtain a transformed image; wherein the transformation operation includes image rotation and / or image flipping;
[0024] The transformed images are respectively input into the multiple deep learning models, the transformed images are encoded, deconvolution is used to expand the encoded feature map to the same size as the iris image, and the enlarged feature map is subjected to an inverse transformation operation opposite to the transformation operation to obtain probability feature maps with the same number as the deep learning model.
[0025] Furthermore, in the process of encoding the iris image and expanding the encoded feature map to the same size as the iris image using deconvolution, the intermediate feature layers are connected in the dimension of channels.
[0026] Furthermore, the method further comprises:
[0027] The maximum connected area is found for the iris effective area, and the maximum connected area is used as the final iris effective area.
[0028] Furthermore, during training, the deep learning model randomly performs data enhancement on image samples in the training data set, and / or the image samples in the training data set include oblique images and / or blurred images;
[0029] Wherein, the data enhancement includes one or more of flipping, translation, noise transformation and elastic deformation;
[0030] The elastic deformation includes: dividing the image sample into sub-blocks of a set size, and performing shape distortion on each sub-block.
[0031] Furthermore, the method further comprises:
[0032] The inner boundary and outer boundary of the iris are fitted according to the inner boundary points and the outer boundary points of the iris effective area.
[0033] Furthermore, the step of inputting the iris image into a pre-trained deep learning model also includes:
[0034] Determine whether the grayscale mean of the iris image is less than a set first grayscale threshold or greater than a set second grayscale threshold, and if so, perform histogram equalization on the iris image;
[0035] The first grayscale threshold is smaller than the second grayscale threshold.
[0036] In a second aspect, the present application provides an iris image segmentation device, the device comprising:
[0037] An image acquisition module, used for acquiring an iris image to be segmented;
[0038] The segmentation module is used to input the iris image into a pre-trained deep learning model, use the iris texture area as the foreground area and the rest as the background area, segment the iris image, and segment an iris effective area for iris feature extraction from the iris image that only includes the iris texture area.
[0039] Furthermore, the segmentation module includes:
[0040] A first processing unit, configured to input the iris image into the deep learning model, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map;
[0041] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area;
[0042] The first judgment unit is used to compare the value of each pixel point on the probability feature map with a set segmentation threshold. If the value of the pixel point in the probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0043] Furthermore, there are multiple deep learning models, and each deep learning model is trained using a different training data set.
[0044] The segmentation module comprises:
[0045] A second processing unit is used to input the iris image into the multiple deep learning models respectively, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image, so as to obtain probability feature maps with the same number as the deep learning model;
[0046] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area;
[0047] The averaging unit is used to calculate the average value of all probability feature maps at the same position to obtain an average probability feature map;
[0048] The second judgment unit is used to compare the value of each pixel point on the average probability feature map with a set segmentation threshold. If the value of the pixel point in the average probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0049] Furthermore, the segmentation module also includes:
[0050] A transformation unit, configured to perform a transformation operation on the iris image to obtain a transformed image; wherein the transformation operation includes image rotation and / or image flipping;
[0051] The third processing unit is used to input the transformed images into the multiple deep learning models respectively, encode the transformed images, use deconvolution to expand the encoded feature map to the same size as the iris image, and perform an inverse transformation operation on the expanded feature map that is opposite to the transformation operation to obtain probability feature maps with the same number as the deep learning model.
[0052] Furthermore, during the execution of the first processing unit, the second processing unit and the third processing unit, the intermediate feature layers are connected in the dimension of the channel.
[0053] Furthermore, the device also includes:
[0054] The post-processing module is used to find the largest connected area in the iris effective area and use the largest connected area as the final iris effective area.
[0055] Furthermore, during training, the deep learning model randomly performs data enhancement on image samples in the training data set.
[0056] Wherein, the data enhancement includes one or more of flipping, translation, noise transformation and elastic deformation;
[0057] The elastic deformation includes: dividing the image sample into sub-blocks of a set size, and performing shape distortion on each sub-block.
[0058] Furthermore, the image samples in the training data set include oblique images and / or blurred images.
[0059] Furthermore, the device also includes:
[0060] The fitting module is used to fit the inner boundary and outer boundary of the iris according to the inner boundary points and outer boundary points of the iris effective area.
[0061] Furthermore, the device also includes:
[0062] The preprocessing module is used to perform scale transformation on the iris image and normalize it to a set size.
[0063] Furthermore, the device also includes:
[0064] A histogram equalization module, used to determine whether the grayscale mean of the iris image is less than a set first grayscale threshold or greater than a set second grayscale threshold, and if so, perform histogram equalization on the iris image;
[0065] The first grayscale threshold is smaller than the second grayscale threshold.
[0066] In a third aspect, the present application provides a computer-readable storage medium for iris image segmentation, comprising a memory for storing processor-executable instructions, wherein the instructions, when executed by the processor, implement the steps of the iris image segmentation method described in the first aspect.
[0067] In a fourth aspect, the present application provides a device for iris image segmentation, characterized in that it includes at least one processor and a memory storing computer executable instructions, and when the processor executes the instructions, the steps of the iris image segmentation method described in the first aspect are implemented.
[0068] This application has the following beneficial effects:
[0069] This application equates the iris segmentation and noise detection problems to a semantic segmentation problem of iris texture, and is a segmentation algorithm for a single iris instance. The method of this application can directly segment the effective iris area, and the segmented effective iris area does not include noise such as eyelids, eyelashes, and light spots. Compared with the multi-stage method of the prior art, the segmentation task becomes simpler and clearer, with higher accuracy, and stronger generalization and robustness.
[0070] This application uses a deep learning method to complete iris segmentation. The parameters of the deep learning model are obtained by learning a large amount of data, rather than the empirical expert system used in the prior art. Due to the diversity of data, the deep learning model learns from diverse samples and has good generalization and robustness to different changes, and performs well under different changes in lighting, angles, etc. It can well solve the problem of iris effective area segmentation in cases of strabismus, squinting, etc. BRIEF DESCRIPTION OF THE DRAWINGS
[0071] Figure 1 This is the effect diagram of positioning the inner and outer circles of the iris;
[0072] Figure 2 This is the effect picture after positioning the inner and outer circles of the iris and the eyelids;
[0073] Figure 3 This is the effect diagram of the noise template;
[0074] Figure 4 This is a flow chart of the iris image segmentation method of the present application;
[0075] Figure 5 is a schematic diagram of an iris image before preprocessing;
[0076] Figure 6 is a schematic diagram of a preprocessed iris image;
[0077] Figure 7 for Figure 5 Schematic diagram of the effective iris area after iris image segmentation;
[0078] Figure 8 Schematic diagram of iris images of various qualities;
[0079] Fig. 9 To use multiple deep learning models Figure 8 Schematic diagram of the result of iris segmentation;
[0080] Fig.10 is a schematic diagram of an iris image before histogram equalization;
[0081] Fig.11 is a schematic diagram of an iris image after histogram equalization;
[0082] Fig.12 This is a schematic diagram of the result of iris segmentation of an iris image after histogram equalization;
[0083] Fig.13 Based on Figure 7 The schematic diagram of the inner boundary of the iris and the outer boundary of the iris shown is fitted in the iris effective area;
[0084] Fig.14 Schematic diagram of the iris image segmentation device of the present application. Detailed implementation manners
[0085] To make the technical problems, technical solutions and advantages to be solved by the present application clearer, the technical solutions of the present application will be clearly and completely described below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Usually, the components of the embodiments of the present application described and shown in the accompanying drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the present application to be protected, but only represents the selected embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.
[0086] The embodiment of the present application provides an iris image segmentation method, as Figure 4 shown, the method includes:
[0087] S100: Obtain an iris image to be segmented.
[0088] The iris image can be collected in real time by an iris collection device, or an iris image collected and stored in advance can be read from a storage space.
[0089] After obtaining the iris image, it can be preprocessed. Through preprocessing, scale transformation is performed on the iris image and normalized to a set size to adapt to the input size of the deep learning model. Exemplarily, the set size can be 512 pixels in both length and width. A specific implementation process of the preprocessing is as follows:
[0090] Judge whether the width (w) and height (h) of the iris image are equal. If w = h, scale the iris image to a size of 512 pixels in both length and width.
[0091] If w > h, calculate the row supplementary boundary height Supplement black boundaries with a width of w and a height of pad1 and a pixel value of 0 to the upper and lower boundaries of the iris image respectively. After supplementation, the width of the iris image is w, and the height is h or h - 2. Then, scale the supplemented iris image to a size of 512 pixels in both length and width.
[0092] If w < h, calculate the row supplementary boundary width The left and right borders of the iris image are supplemented with black borders with a width of pad1, a height of h, and a pixel value of 0, respectively, and the supplemented iris image has a width of w or w-2 and a height of h. Then, the supplemented iris image is scaled to a size of 512 pixels in both length and width.
[0093] In one example, the iris image before preprocessing is as follows: Figure 5 As shown in Figure 2, the preprocessed iris image is Figure 6 shown.
[0094] S200: Input the iris image into a pre-trained deep learning model, use the iris texture area as the foreground area and the rest as the background area, segment the iris image, and segment an iris effective area for iris feature extraction from the iris image that only includes the iris texture area.
[0095] This application uses a deep learning method to solve the problem of iris effective area positioning in iris recognition. The iris image is segmented based on the iris texture, and the iris texture is regarded as a category as the foreground area, and the rest of the non-iris part is regarded as a category, such as the eyelids, eyelashes, the area of the reflected light spot, and other areas, which are uniformly regarded as the non-iris area as the background area. Figure 5 The effective iris area after iris image segmentation is shown as Figure 7 The segmented iris effective area only includes the iris texture area, which is used for iris feature extraction in iris recognition.
[0096] The existing iris image segmentation is a multi-stage method, which requires locating the inner and outer boundaries of the iris on the iris image, then performing iris noise template detection, and finally segmenting the iris effective area. The iris effective area cannot be directly segmented in one step. Each step has a great impact on the accuracy of iris recognition, and the steps are relatively cumbersome and complicated.
[0097] This application equates the iris segmentation and noise detection problems to a semantic segmentation problem of iris texture, and is a segmentation algorithm for a single iris instance. The method of this application can directly segment the iris effective area that only includes the iris texture area, and the segmented iris effective area does not include noise such as eyelids, eyelashes, and light spots. Compared with the multi-stage method of the prior art, the segmentation task becomes simpler and clearer, with higher accuracy, and stronger generalization and robustness.
[0098] Existing iris image segmentation methods require iris noise detection. The types of iris noise include eyelid occlusion, eyelash occlusion, and light spot occlusion. Each type of noise requires detection and segmentation using digital image processing methods. The methods for detecting eyelids, eyelashes, and light spots are also inconsistent. Not only are the steps cumbersome, but traditional noise detection methods are easily affected by factors such as lighting and angles, and the algorithm generalization is poor. This application directly implements iris segmentation through a deep learning method without the need for iris noise detection, avoiding the problems of various noise detection methods being cumbersome, easily affected by lighting and angles, and having poor generalization.
[0099] Traditional methods for locating the inner and outer boundaries of the iris are generally based on the gradient of the image (such as the calculus detection operator based on the gradient amplitude and the Hough transform based on the edge gradient binarization, etc.). The inner and outer boundaries of the iris are determined by calculating the gradient of the image, and its generalization is poor. Moreover, for strabismus and blurred iris images, the inner and outer boundaries of the iris cannot be accurately located, and image quality needs to be controlled. Traditional iris and eyelash detection methods use expert systems, and most of them use methods such as threshold segmentation for image segmentation or positioning. This method has poor generalization, is greatly affected by the environment, and the algorithm performs poorly on large-scale data.
[0100] This application uses a deep learning method to complete iris segmentation. The parameters of the deep learning model are obtained by learning a large amount of data, rather than the empirical expert system used in the prior art. Due to the diversity of data, the deep learning model learns from diverse samples and has good generalization and robustness to different changes, and performs well under different changes in lighting, angles, etc. It can well solve the problem of iris effective area segmentation in cases of strabismus, squinting, etc.
[0101] Traditional iris recognition usually has many limitations in terms of user cooperation and image conditions, which seriously affect the widespread application of iris recognition. In recent years, many non-cooperative or low-restriction iris recognition methods have also emerged, such as long-distance iris recognition, iris recognition in motion, iris recognition without perception and cooperation, iris recognition under dynamic image environment changes, and iris recognition on mobile devices. The iris area collected under these environments usually contains a variety of noises, such as occlusion caused by eyelids or eyelashes, mirror reflection, strabismus, blur, etc.
[0102] For the above-mentioned iris recognition scenario containing a lot of noise, the present application improves the generalization ability of iris segmentation through a deep learning method, can be well applied to the above-mentioned noisy scenario, realizes iris recognition on noisy images, and can perform effective and robust iris segmentation.
[0103] The deep learning model used in this application may be a convolutional neural network, and the aforementioned S200 includes:
[0104] S210: Input the iris image into the deep learning model, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map.
[0105] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area.
[0106] In order to better combine semantic features and spatial features, in this step, the intermediate feature layers can be connected in the channel dimension to perform feature enhancement.
[0107] In one example, the specific implementation of S210 of the present application includes:
[0108] 1. Input an iris image of size 512x512 and 3 channels to the deep learning model.
[0109] 2. For the above iris image, use a convolution kernel of size 7x7, a step size of 2, a convolution operation with 3 pixels padded on the edge, and perform BN and ReLU operations to obtain a 64-channel feature map f0 of size 256x256.
[0110] 3. Perform a convolution operation on the above feature map with a convolution kernel of size 3x3, a step size of 2, and a 1-pixel padding on the edge to obtain a 64-channel feature map A1 of size 128x128.
[0111] 4. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 64-channel feature map of size 128x128.
[0112] 5. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a 64-channel feature map B1 of size 128x128.
[0113] 6. Add feature map A1 and feature map B1 pixel by pixel on each channel to obtain a 64-channel feature map f1 with a size of 128x128
[0114] 7. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 64-channel feature map of size 128x128.
[0115] 8. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a 64-channel feature map B2 with a size of 128x128.
[0116] 9. Add the feature map f1 and the feature map B2 pixel by pixel on each channel to obtain a 64-channel feature map f2 of size 128x128.
[0117] 10. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a 1-pixel edge padding, and perform BN and ReLU operations to obtain a 64-channel feature map of size 128x128.
[0118] 11. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a 64-channel feature map B3 with a size of 128x128.
[0119] 12. Add feature map f2 and feature map B3 pixel by pixel on each channel to obtain a 64-channel feature map f3 with a size of 128x128.
[0120] 13. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 2, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of 128 channels with a size of 64x64.
[0121] 14. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B4 of size 64x64 and 128 channels.
[0122] 15. Use a convolution operation with a convolution kernel of size 1x1, a step size of 1, and fill the edges with 0 pixels for the feature map f3 to obtain a feature map A3 of size 64x64 and 128 channels.
[0123] 16. Add feature map A3 and feature map B4 pixel by pixel on each channel to obtain a 128-channel feature map f4 of size 64x64.
[0124] 17. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of 128 channels with a size of 64x64.
[0125] 18. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B5 of size 64x64 and 128 channels.
[0126] 19. Add the feature map f4 and the feature map B5 pixel by pixel on each channel to obtain a 128-channel feature map f5 of size 64x64.
[0127] 20. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 64x64 with 128 channels.
[0128] 21. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B6 of size 64x64 and 128 channels.
[0129] 22. Add the feature map f5 and the feature map B6 pixel by pixel on each channel to obtain a 128-channel feature map f7 of size 64x64.
[0130] 23. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 64x64 with 128 channels.
[0131] 24. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B7 of size 64x64 and 128 channels.
[0132] 25. Add the feature map f7 and the feature map B7 pixel by pixel on each channel to obtain a 128-channel feature map f8 of size 64x64.
[0133] 26. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 2, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0134] 27. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B8 of size 32x32 and 256 channels.
[0135] 28. Use a convolution operation with a convolution kernel of size 1x1, a stride of 1, and fill the edges with 0 pixels for the feature map f8 to obtain a feature map A8 of size 32x32 and 256 channels.
[0136] 29. Add feature map A8 and feature map B8 pixel by pixel on each channel to obtain a feature map f9 with a size of 32x32 and 256 channels.
[0137] 30. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0138] 31. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B9 of size 32x32 and 256 channels.
[0139] 32. Add feature map f9 and feature map B9 pixel by pixel on each channel to obtain a feature map f with a size of 32x32 and 256 channels. 10 .
[0140] 33. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0141] 34. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 32x32 with 256 channels. 10 .
[0142] 35. The feature map f 10 And feature map B 10 Add pixel by pixel on each channel to get a 256-channel feature map f with a size of 32x32. 11 .
[0143] 36. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0144] 37. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 32x32 with 256 channels. 11 .
[0145] 38. The feature map f 11 And feature map B 11 Add pixel by pixel on each channel to get a 256-channel feature map f with a size of 32x32. 12 .
[0146] 39. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0147] 40. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 32x32 with 256 channels. 12 .
[0148] 41. The feature map f 12 And feature map B 12 Add pixel by pixel on each channel to get a 256-channel feature map f with a size of 32x32. 13 .
[0149] 42. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of size 32x32 with 256 channels.
[0150] 43. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 32x32 with 256 channels. 13 .
[0151] 44. The feature map f 13 And feature map B 13 Add pixel by pixel on each channel to get a 256-channel feature map f with a size of 32x32. 14 .
[0152] 45. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 2, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of 512 channels with a size of 16x16.
[0153] 46. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 16x16 with 512 channels. 14 .
[0154] 47. For the feature map f 14 Using a convolution kernel of size 1x1, a stride of 1, and a convolution operation with 0 pixels padded on the edges, we get a feature map A of size 16x16 with 512 channels. 14 .
[0155] 48. Feature map A 14 And feature map B14 Add pixel by pixel on each channel to get a 512-channel feature map f with a size of 16x16. 15 .
[0156] 49. For the above feature map, use a convolution operation with a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map of 512 channels with a size of 16x16.
[0157] 50. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform a BN operation to obtain a feature map B of size 16x16 with 512 channels. 15 .
[0158] 51. The feature map f 15 And feature map B 15 Add pixel by pixel on each channel to get a 512-channel feature map f with a size of 16x16. 16 .
[0159] At this point, the process of encoding the iris image is completed. The subsequent process is to use deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map.
[0160] 52. The feature map f 16 Use the nearest neighbor interpolation to perform 2x interpolation to obtain a 512-channel feature map f with a size of 32x32. 17 .
[0161] 53. 17 and f 14 Connect in the channel dimension to get a feature map f with 768 channels of size 32x32 18 .
[0162] 54. For f 18 A convolution operation with a convolution kernel of size 3x3, a stride of 1, and 1 pixel padding on the edge is performed, and BN and ReLU operations are performed to obtain a feature map of size 32x32 with 256 channels.
[0163] 55. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a feature map f of size 32x32 with 256 channels. 19 .
[0164] 56. The feature map f 19 Use the nearest neighbor interpolation to perform 2x interpolation to obtain a feature map f with a size of 64x64 and 256 channels 20 .
[0165] 57. 20 Connect f8 with f in the channel dimension to obtain a feature map f with a size of 64x64 and 384 channels. 21 .
[0166] 58. For f 21 A convolution operation with a convolution kernel of size 3x3, a stride of 1, and 1 pixel padding on the edge is performed, and BN and ReLU operations are performed to obtain a feature map of 128 channels with a size of 64x64.
[0167] 58. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 128-channel feature map f with a size of 64x64. 22 .
[0168] 60. The feature map f 22 Use the nearest neighbor interpolation to perform 2x interpolation to obtain a 128-channel feature map f with a size of 128x128. 23 .
[0169] 61. 23 Connect f3 with f in the channel dimension to obtain a 192-channel feature map f with a size of 128x128 24 .
[0170] 62. For f 24 A convolution operation with a convolution kernel of size 3x3, a stride of 1, and 1 pixel padding on the edge is performed, and BN and ReLU operations are performed to obtain a 64-channel feature map of size 128x128.
[0171] 63. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 64-channel feature map f with a size of 128x128. 25 .
[0172] 64. The feature map f 25 Use the nearest neighbor interpolation to perform 2x interpolation to obtain a 128-channel feature map f with a size of 256x256. 26 .
[0173] 65. 26 Connect f0 with f0 in the channel dimension to obtain a 192-channel feature map f with a size of 256x256 27 .
[0174] 66. To f 27A convolution operation with a convolution kernel of size 3x3, a stride of 1, and 1 pixel padding on the edge is performed, and BN and ReLU operations are performed to obtain a 64-channel feature map of size 256x256.
[0175] 67. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, and a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 64-channel feature map f with a size of 256x256. 28 .
[0176] 68. The feature map f 28 Use the nearest neighbor interpolation to perform 2x interpolation to obtain a 128-channel feature map f with a size of 512x512. 29 .
[0177] 69. For f 29 Use a convolution kernel of size 3x3, a stride of 1, a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 16-channel feature map of size 512x512.
[0178] 70. For the above feature map, use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a 16-channel feature map f with a size of 512x512. 30 .
[0179] 71. For f 30 Use a convolution kernel of size 3x3, a step size of 1, a convolution operation with 1 pixel padding on the edge, and perform BN and ReLU operations to obtain a single-channel feature map f of size 512x512. 31 .
[0180] 72. Finally, 31 Using the softmax operation, we can get a probability feature map P of size 512x512. The value of each pixel on the probability feature map P represents the probability that the pixel at the same position corresponding to the original iris image is the valid iris area.
[0181] S220: Compare the value of each pixel point on the probability feature map with the set segmentation threshold. If the value of the pixel point on the probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is determined to belong to the iris effective area.
[0182] Assuming that 0.5 is used as the segmentation threshold, if the value p(x,y) of a pixel at any position (x,y) in the probability feature map P is greater than 0.5, it indicates that the pixel at position (x,y) on the iris image belongs to the effective iris area.
[0183] In different iris recognition systems, the quality of the collected iris images varies greatly (e.g. Figure 8 As shown, including iris images of various qualities), a positioning and segmentation algorithm with good generalization and strong robustness is extremely important. This application uses a multi-model integration method to set the deep learning model to multiple (for example, 5) and determines the iris effective area by the average value of the probability feature map output by multiple deep learning models, so that the iris segmentation result is more accurate. Accordingly, the S200 includes:
[0184] S230: Input the iris image into multiple deep learning models respectively, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image, so as to obtain the same number of probability feature maps as the deep learning model.
[0185] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area.
[0186] Each of the multiple deep learning models is trained using a different training data set. The different training data sets may be different independent data sets established, or may be a total data set divided into multiple different training data sets.
[0187] Assuming that the number of deep learning models is 5, the same iris image is processed by 5 deep learning models respectively, and 5 probability feature maps can be obtained.
[0188] S240: Calculate the average value of all probability feature maps at the same position to obtain an average probability feature map.
[0189] S250: Compare the value of each pixel point on the average probability feature map with the set segmentation threshold. If the value of the pixel point on the average probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0190] The above-mentioned multiple deep learning models of this application are used to Figure 8 The iris images of various qualities are segmented, and the results are as follows Fig. 9 As shown, it can be seen that the result is relatively accurate.
[0191] In order to further improve the accuracy of the iris segmentation results, the present application adopts a data enhancement operation during the processing of the iris image by the deep learning model (i.e., during the testing process). Specifically, the S200 further includes:
[0192] S231: Performing a transformation operation on the iris image to obtain a transformed image.
[0193] The transformation operation includes image rotation and / or image flipping. In one example, the transformation operation is upside down flipping.
[0194] S232: Input the transformed images into multiple deep learning models respectively, encode the transformed images, use deconvolution to expand the encoded feature map to the same size as the iris image, and perform an inverse transformation operation on the expanded feature map that is opposite to the transformation operation to obtain probability feature maps with the same number as the deep learning model.
[0195] Assuming that the number of deep learning models is 5, the same transformed image is processed by 5 deep learning models respectively, and 5 probability feature maps can be obtained. Since the input is a transformed image, the output feature map needs to be inversely transformed to correspond to the original iris image.
[0196] The five probability feature maps of the iris image after being processed by the five deep learning models are denoted as Seg i (i=0,1,2,3,4), the five probability feature maps after the transformed image is processed by the five deep learning models are denoted as Seg i (i=5,6,7,8,9).
[0197] Next, in S240, the average value is calculated to obtain the average probability characteristic graph Seg, the second probability characteristic graph and the third probability characteristic graph.
[0198]
[0199] After the above-mentioned probability feature map or average probability feature map is used to determine the effective iris area, it can be directly used as the final result. Of course, the effective iris area can also be post-processed, and the post-processing method includes:
[0200] S300: searching for the largest connected area in the iris effective area, and taking the largest connected area as the final iris effective area.
[0201] By finding the largest connected area, the disconnected abnormal points can be removed and the segmentation result with the largest connected area can be retained. In this application, for each pixel point segmented, the 4-neighborhood maximum connected area method can be used to find the largest connected area, and the pixel point of the largest connected area is used as the final iris segmentation result.
[0202] In the segmentation of the iris area, it is more difficult to segment iris images that are flipped upside down, squinted, and squint. To address the above problems, the deep learning model of the present application randomly performs data enhancement on image samples in the training data set during training.
[0203] Specifically, the image samples in the training data set are randomly flipped, translated, and noise transformed to make the data diverse. In addition, the elastic deformation method can be used to divide the image samples into sub-blocks of a set size (for example, 10*10 size), and each sub-block is distorted. The deep learning model trained by the above data enhancement method can well solve the segmentation problem of iris images such as strabismus and squinting, which helps to improve the generalization and accuracy of the deep learning model.
[0204] In the prior art, it is not possible to locate the boundary of strabismus and blurred iris images well, and thus it is not possible to obtain an accurate iris effective area. To solve this problem, the present application makes the image samples in the training data set include strabismus images and / or blurred images, which can better solve the segmentation problem of strabismus and blurred iris images and have a more accurate segmentation result.
[0205] In the process of iris image segmentation, there are cases where the iris image with weak illumination or overexposure cannot be segmented or the segmentation result is incorrect, such as Fig.10 To solve the above problems, the method of the present application further includes:
[0206] S110: Determine whether the grayscale mean of the iris image is less than a set first grayscale threshold or greater than a set second grayscale threshold. If so, perform histogram equalization on the iris image.
[0207] The first grayscale threshold is smaller than the second grayscale threshold.
[0208] The present application determines the brightness of the iris image. For example, the value of the first grayscale threshold may be 50, and the value of the second grayscale threshold may be 200. When the grayscale mean of the iris image is less than 50 or greater than 200, the iris image needs to be histogram equalized. The iris image after histogram equalization can be better segmented.
[0209] The iris image before histogram equalization is as follows Fig.10 As shown in the figure, the iris image after histogram equalization is as follows Fig.11 As shown in the figure, the result of segmentation of the iris image after histogram equalization is as follows Fig.12 As shown, it can be seen that the result is relatively accurate.
[0210] In this application, after determining the effective iris area, the following steps may also be included:
[0211] S300: fitting the inner boundary and outer boundary of the iris according to the inner boundary points and the outer boundary points of the effective area of the iris.
[0212] During fitting, the inner and outer boundaries of the iris are fitted as an ellipse. The ellipse described in the present application also includes a perfect circle, because a perfect circle can also be regarded as a special ellipse in terms of expression.
[0213] The general equation of an ellipse is:
[0214] Ax 2 +Bxy+Cx 2 +Dx+Ey+F=0
[0215] Where A, B, C, D, E, and F are the parameters to be determined. Substituting the inner boundary points / outer boundary points of the effective area of the iris into the above equation, the expression of the ellipse can be obtained, and the inner boundary / outer boundary of the iris can be obtained.
[0216] An example of the fitted iris inner boundary and iris outer boundary is as follows Fig.13 As shown, the iris inner boundary and the iris outer boundary can be used to locate the iris, estimate the actual size and position of the iris in the image, etc., and can also be used to correct the iris image.
[0217] The present application also provides an iris image segmentation device. Fig.14 As shown, the device comprises:
[0218] The image acquisition module 100 is used to acquire an iris image to be segmented.
[0219] The segmentation module 200 is used to input the iris image into a pre-trained deep learning model, use the iris texture area as the foreground area and the rest as the background area, segment the iris image, and segment an iris effective area that only includes the iris texture area from the iris image for iris feature extraction.
[0220] This application equates the iris segmentation and noise detection problems to a semantic segmentation problem of iris texture, and is a segmentation algorithm for a single iris instance. The method of this application can directly segment the effective iris area, and the segmented effective iris area does not include noise such as eyelids, eyelashes, and light spots. Compared with the multi-stage method of the prior art, the segmentation task becomes simpler and clearer, with higher accuracy, and stronger generalization and robustness.
[0221] This application uses a deep learning method to complete iris segmentation. The parameters of the deep learning model are obtained by learning a large amount of data, rather than the empirical expert system used in the prior art. Due to the diversity of data, the deep learning model learns from diverse samples and has good generalization and robustness to different changes, and performs well under different changes in lighting, angles, etc. It can well solve the problem of iris effective area segmentation in cases of strabismus, squinting, etc.
[0222] The deep learning model of the present application may be a convolutional neural network, and accordingly, the segmentation module includes:
[0223] The first processing unit is used to input the iris image into the deep learning model, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map.
[0224] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area.
[0225] The first judgment unit is used to compare the value of each pixel point on the probability feature map with the set segmentation threshold. If the value of the pixel point on the probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0226] The number of deep learning models used in this application can be multiple, and each deep learning model is trained by a different training data set. Accordingly, the segmentation module includes:
[0227] The second processing unit is used to input the iris image into multiple deep learning models respectively, encode the iris image, and use deconvolution to expand the encoded feature map to the same size as the iris image, so as to obtain the same number of probability feature maps as the deep learning model.
[0228] The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area.
[0229] The averaging unit is used to calculate the average value of all probability feature maps at the same position to obtain the average probability feature map.
[0230] The second judgment unit is used to compare the value of each pixel point on the average probability feature map with the set segmentation threshold. If the value of the pixel point in the average probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
[0231] Furthermore, the segmentation module also includes:
[0232] The transformation unit is used to perform a transformation operation on the iris image to obtain a transformed image; wherein the transformation operation includes image rotation and / or image flipping.
[0233] The third processing unit is used to input the transformed images into multiple deep learning models respectively, encode the transformed images, use deconvolution to expand the encoded feature map to the same size as the iris image, and perform an inverse transformation operation on the expanded feature map that is opposite to the transformation operation to obtain probability feature maps with the same number as the deep learning model.
[0234] Furthermore, during the execution of the first processing unit, the second processing unit and the third processing unit, the intermediate feature layers are connected in the dimension of the channel.
[0235] As an improvement of the embodiment of the present application, the device further includes:
[0236] The post-processing module is used to find the largest connected area in the iris effective area and use the largest connected area as the final iris effective area.
[0237] During training, the deep learning model of the present application can randomly perform data augmentation on image samples in the training data set.
[0238] Specific data enhancement includes one or more of flipping, translation, noise transformation and elastic deformation.
[0239] The elastic deformation includes: dividing the image sample into sub-blocks of a set size and performing shape distortion on each sub-block.
[0240] In order to improve the segmentation effect of strabismus and blurred iris images, the image samples in the training data set of the present application include strabismus images and / or blurred images.
[0241] As another improvement of the embodiment of the present application, the device further includes:
[0242] The fitting module is used to fit the inner boundary and outer boundary of the iris according to the inner boundary points and outer boundary points of the iris effective area.
[0243] and / or;
[0244] The preprocessing module is used to scale the iris image and normalize it to a set size.
[0245] and / or;
[0246] The histogram equalization module is used to determine whether the grayscale mean of the iris image is less than a set first grayscale threshold or greater than a set second grayscale threshold. If so, histogram equalization is performed on the iris image.
[0247] The first grayscale threshold is smaller than the second grayscale threshold.
[0248] The device provided in the above embodiment, its implementation principle and the technical effect produced correspond to the embodiment of the above method one by one. For the sake of brief description, the part not mentioned in the embodiment of the device can refer to the corresponding content in the embodiment of the above method. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working process of the modules and units described in the device can refer to the corresponding process in the embodiment of the above method, and will not be repeated here.
[0249] The iris image segmentation method described in the above embodiments provided by the present application can implement the business logic through a computer program and record it on a storage medium, and the storage medium can be read and executed by a computer to achieve the effect of the scheme described in the method embodiment of this specification. Therefore, the embodiment of the present application also provides a computer-readable storage medium for iris image segmentation, including a memory for storing processor-executable instructions, and when the instructions are executed by the processor, the steps of the iris image segmentation method including the above embodiments are implemented.
[0250] The storage medium may include a physical device for storing information, which is usually a medium that digitizes the information and then stores it in an electrical, magnetic or optical manner. The storage medium may include: a device that stores information in an electrical energy manner, such as various memories, such as RAM, ROM, etc.; a device that stores information in a magnetic energy manner, such as a hard disk, a floppy disk, a magnetic tape, a magnetic core memory, a magnetic bubble memory, a USB flash drive; a device that stores information in an optical manner, such as a CD or a DVD. Of course, there are other readable storage media, such as quantum memory, graphene memory, etc.
[0251] The storage medium described above may also include other implementation methods according to the description of the method embodiment. The implementation principle and technical effects produced by this embodiment are the same as those of the aforementioned method embodiment. For details, please refer to the description of the relevant method embodiment, and no further description will be given here.
[0252] The embodiment of the present application also provides a device for iris image segmentation, which may be a separate computer, or may include an actual operating device using one or more of the methods or one or more of the embodiments of the present specification. The device for iris image segmentation may include at least one processor and a memory storing computer executable instructions, and when the processor executes the instructions, it implements any one or more of the steps of the iris image segmentation method described above.
[0253] The device described above may also include other implementation methods according to the description of the method embodiment. The implementation principle and technical effects produced by this embodiment are the same as those of the aforementioned method embodiment. For details, please refer to the description of the relevant method embodiment, and no further description will be given here.
[0254] Finally, it should be noted that the above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the above-mentioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the above-mentioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical features therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.
Claims
1. A method for segmenting an iris image, characterized in that: The method comprises: Obtain an iris image to be segmented; The iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction that only includes the iris texture area is segmented from the iris image.
2. The iris image segmentation method according to claim 1, characterized in that: The iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction including only the iris texture area is segmented from the iris image, including: Inputting the iris image into the deep learning model, encoding the iris image, and using deconvolution to expand the encoded feature map to the same size as the iris image to obtain a probability feature map; The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area; The value of each pixel point on the probability feature map is compared with the set segmentation threshold. If the value of the pixel point in the probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
3. The iris image segmentation method according to claim 1, characterized in that: There are multiple deep learning models, and each deep learning model is trained using a different training data set. The iris image is input into a pre-trained deep learning model, the iris texture area is used as the foreground area, and the rest are used as the background area, the iris image is segmented, and an iris effective area for iris feature extraction including only the iris texture area is segmented from the iris image, including: Inputting the iris image into the multiple deep learning models respectively, encoding the iris image, and using deconvolution to expand the encoded feature map to the same size as the iris image, so as to obtain the same number of probability feature maps as the deep learning model; The value of each pixel point on the probability feature map represents the probability that the corresponding pixel point at the same position on the iris image belongs to the iris effective area; Calculate the average value of all probability feature maps at the same position to obtain the average probability feature map; The value of each pixel point on the average probability feature map is compared with the set segmentation threshold. If the value of the pixel point in the average probability feature map is greater than the segmentation threshold, the corresponding pixel point at the same position on the iris image is judged to belong to the iris effective area.
4. The iris image segmentation method according to claim 3, characterized in that: The calculation of the average value of all probability feature maps at the same position also includes: Performing a transformation operation on the iris image to obtain a transformed image; wherein the transformation operation includes image rotation and / or image flipping; The transformed images are respectively input into the multiple deep learning models, the transformed images are encoded, deconvolution is used to expand the encoded feature map to the same size as the iris image, and the enlarged feature map is subjected to an inverse transformation operation opposite to the transformation operation to obtain probability feature maps with the same number as the deep learning model.
5. The iris image segmentation method according to any one of claims 2 to 4, characterized in that: In the process of encoding the iris image and using deconvolution to expand the encoded feature map to the same size as the iris image, the intermediate feature layer is connected in the dimension of the channel.
6. The iris image segmentation method according to any one of claims 1 to 4, characterized in that: The method further comprises: The maximum connected area is found for the iris effective area, and the maximum connected area is used as the final iris effective area.
7. The iris image segmentation method according to claim 1, characterized in that: When the deep learning model is being trained, data enhancement is randomly performed on image samples in the training data set, and / or the image samples in the training data set include oblique images and / or blurred images; Wherein, the data enhancement includes one or more of flipping, translation, noise transformation and elastic deformation; The elastic deformation includes: dividing the image sample into sub-blocks of a set size, and performing shape distortion on each sub-block.
8. The iris image segmentation method according to claim 1, characterized in that: The method further comprises: The inner boundary and outer boundary of the iris are fitted according to the inner boundary points and the outer boundary points of the iris effective area.
9. The iris image segmentation method according to claim 1, characterized in that: The step of inputting the iris image into a pre-trained deep learning model also includes: Determine whether the grayscale mean of the iris image is less than a set first grayscale threshold or greater than a set second grayscale threshold, and if so, perform histogram equalization on the iris image; The first grayscale threshold is smaller than the second grayscale threshold.
10. An iris image segmentation device, characterized in that: The device comprises: An image acquisition module, used for acquiring an iris image to be segmented; The segmentation module is used to input the iris image into a pre-trained deep learning model, use the iris texture area as the foreground area and the rest as the background area, segment the iris image, and segment an iris effective area for iris feature extraction from the iris image that only includes the iris texture area.
11. A computer-readable storage medium for iris image segmentation, characterized in that: It comprises a memory for storing processor executable instructions, wherein when the instructions are executed by the processor, the steps of the iris image segmentation method according to any one of claims 1 to 9 are implemented.
12. A device for iris image segmentation, characterized in that: The method comprises at least one processor and a memory storing computer executable instructions, wherein the processor implements the steps of the iris image segmentation method according to any one of claims 1 to 9 when executing the instructions.