International Journal of Interactive Mobile Technologies (iJIM) – eISSN: 1865-7923 – Vol 16 No 24 (2022) Paper—Identification Based on Iris Detection Technique Identification Based on Iris Detection Technique https://doi.org/10.3991/ijim.v16i24.36433 Shaimaa Hameed Shaker1(), Farah Qais Al-Kalidi2, Raheem Ogla1 1 University of Technology, Baghdad, Iraq 2 Almustansriya University, Baghdad, Iraq shaimaa.h.shaker@uotechnology.edu.iq Abstract—Iris-biometrics are an alternative way of authenticating and iden- tifying a person because biometric identifiers are unique to people. This paper introduces a method aims to efficient human identification by enhanced iris de- tection method within acceptable time. After preparing various type of images, then perform a series of pre-processing steps and standardize them, after that use Uni-Net learning, so identify the human by Navie-Bays method is the last step based on the output of Uni-Net which is role as feature extractor for the iris part and another sub-net for non-iris part that may involve identification-outcome. The outcome of this method looked good compared to some high-level methods, so, was accuracy-rate 9855, 99.25, and 99.81 for CASIA-v4, ITT-Delhi, and MMU-database respectively. Also, this paper introduces a method of iris recog- nition using CNN model which is improved the preprocessed patterns that to- gether from dataset applied some procedures to develop them based on tech- niques of equalization and acclimate contrast ones. After that characteristic ex- tracted and classified using CNN that comprises of 10 layers with back-propaga- tion schema and adjusted moment evaluation Adam-optimizer for modernize weights. The overall accuracy was 95.31% with utilization time 17.58 (mints) for training-model. Keywords—identification, biological data measures, deep structured-learning, UniNet, fully connected neural network, convolutional deep neural network, Eyeiris detection 1 Introduction Authentication, identification, and detection are progressed extensively depending on biometric systems, also used to elevate the privacy and security involved[1, 2]. Bi- ometrical measures define as “physical characteristics measurements that are utilized to discriminate a person” [3]. So, these measures use physical and behavioral features to detect and distinguish persons, for this reason, systems of biometric measures are used as an efficient solution of security where the properties cannot be faked, stolen, lost, or forgotten [4]. Recently, demand has increased for improving daily life security in the digitalization direction in order to enhance reliability and intelligent identification systems using biometrics [5]. Traditional eye-iris detection systems are based on the least space between a person and a pictural-device which is less than 1m. So, these 154 http://www.i-jim.org https://doi.org/10.3991/ijim.v16i24.36433 mailto:shaimaa.h.shaker@uotechnology.edu.iq Paper—Identification Based on Iris Detection Technique systems are based on high user cooperation but the drawback is a person has to pause and look at the sensor-device [6]. The permanence of eye-iris attributes for a person along the living makes the eye-iris detection the preferred one for individual identifica- tion [7, 8]. Biometric measures often consist of sequential as: Registration, pre dealing, mapping, and decision-making processes, so numerous functions and strategies for each phase, chiefly for the concluding phase that it is essential for delivering outcome- result[9]. Although the person recognition and identification system are one of the most reliable biometric techniques, the poor factor of the captured images is the main reason for the damage to the performance of this system and elevated False Rejection Rates (FRR) thus its failure to accurately identify the iris of each person [10, 11]. The typical nets of neural technique trouble are that it requires a lot of predealing out and modification to provide high-quality fallout in certified data, also no agreement to have an active answer on all biometrics-measures or on another dataset for the same bio-system. To defeat this trouble, the investigators barely locate characteristics then deep-NN has reached good quality outcomes on numerous standard data. Latest time, deep CNN-supported learning approaches have acquired marvelous achievement in scope of computer vision and pattern recognition applications. Deep learning-based tactics have shown preponderate outcome for a variety of tribulations related to the assignment and catching biometrics[12]. Convolutional neural network (CNN) is a cat- egory of non-natural intelligence and has the ability to learn and extract the features automatically as a human, whereas the classical neural network has the inability to learn features automatically [5, 12, 13]. This research paper includes a work which is used an extra preprocessing step before deep learning is used to get time consuming and rapid enhancement of iris image detection. There is a lot of a number of related works which depend on deep learning neural networks for iris detection such as: In 2017, Zhao and Kumar proposed a perfect eye-iris recall based on totally learned related character- istics[14]. Again in 2017, Al‑Waisy, and et. al introduce an eye-iris detection system by means of CNN and soft-max classifier as a union-model based on multi data-sets as a trail to answer over-fitting problem, they got some results with accelerate but needing to be enhanced [15]. In 2018, Shi, and et. al they attempted to work out the trouble of unfinished and not apparent images that is utilized for eye-iris classification so have trained a normalization part of iris using convolution neural network, but face the over- fitting problem [16]. Liu, and et. al, in 2019 trained deep learning methods after using Gaussian, triangular-fuzzy-average, and triangular-fuzzy-median smoothing filters to pre-deal with image by vague the area outside the margin to upgrade the value of signal- to-noise scales so speeded up task of convergence and raises the recognition-rate [17, 18]. In 2020, Omran and Al-Shemmary used a deep convolutional neural network to detect iris by classifying the main features of the Indian-Institute-Technology-Delhi (IITD) eye-iris data after extracting them [19]. Anca and Ioan 2020 introduced a SURF- descriptors and quality characteristics utilized in eye-iris detection. They got improved results obtained by applying the method and showed that the schema can be utilized as a sift to minimize the analysis of eye-iris identification [20, 21].In 2021, V. Conti1 and et.al| presented multiple construct biometric systems via eye-iris with eye-retina together. So, the work solved arrangement and recall method usually implement in iJIM ‒ Vol. 16, No. 24, 2022 155 https://www.researchgate.net/scientific-contributions/Ming-Liu-2160518944?_sg%5B0%5D=syPOX9uqEi2Hf3n6YRiEK7Lk8GrWklWPLNkNdM11rjF98fAhWxpSpC25VkBqolCHCf-hpEs.TElRW79JlB7MRjxOA0oQgoROZmUfuEq2eJkERKimRkkjrYf1vbQVmivObcLjKt8c0YtRbaHQFiYQWSZL3BhOfA&_sg%5B1%5D=PHAvB4mZ2wBHEU1jyesrwe0JhS6tkhZ09GLq-71TpEPhREfdRSXkyu2GWujAoCsozwgzel8.DNqG--vkFqY-R33Ane0ZU5u68F6swGwOmNg-tvubQxU2GdkNA7m-xqb32N27HyZVH5EZt8JcI14-dXananGCqA Paper—Identification Based on Iris Detection Technique computerized bioinformatics; such as attributes of main features are selected inde- pendently from all eye-iris so as all eye-retina, and the combination is achieved by contrast mark using the Levenshtein distance [22]. The research-paper aims to rely on iris detection for accurate and optimal identification of the person in the least time and cost. This paper is structured as: After the current section, the description of both the two-methodidentification-based iris detection model utilized UniNet and a method of iris recognition model utilized deep learning technology-based CNN are represented in 2nd section. While the 3rd section discusses the experimental results. Some conclusions are shown in the last section. 1.1 Difficulties of eye-identification technique Pre progressing is an essential action in the eye recognition system and involves the exactness of identical. Some of the most important actions are eye-iris pinpoint ion and eye-iris image excellence valuation. Iris-pinpointing is discovery of the internal and external borders of the eye-iris. The pinpoint ion schema aspires for rapid and precise willpower of the eye-iris borders. So accurate schemes need a lengthy time to situate iris [23, 24]. Start step is the pinpointing which means detecting the eye-pupil, which is the black rounded component enclosed by iris tissues. Pupil-edges is can be detected easily from the binarized image because the eye-pupile is the biggest darkness by using a suitable threshold on the intensity image. However, the trouble of binarization hap- pens if someone having very gloomy pupil, and the pinpointing does not succeed for this reason required to decide on a reasonable threshold value. Again, sometimes the pupil is not an ideal perimeter [25].At the same time, Little excellent eye-images con- tain deprived illumination defocus haze, of-angle, and serious occlusion, So This does not give any positive result even if the best hashing algorithms are used [26, 27]. Irises may be catched in diverse dimension same as the iris of the same one may alter for the reason that difference of the lighting. Therefore, need Normalization that regulates data to reduce redundancy and separate a database into records and with relations between the records. The problems are that the eye-image has problems of illumination. So, image enhancement using histogram equalization and some filtering to noise elimina- tion. The problem with histogram equalization is that shaped unwanted results when operated on iris-images with low color depth [25]. High dimensional troubles are be- coming more and more ordinary. So, feature extraction is the next action of identifica- tion by bio measures that generates feature vectors corresponding to human being iris images and to matching based on a number of distance measures. One of these difficul- ties in attributes-based identification of iris is matching performance is prejudiced by a lot of factors in the feature selection process. Besides that, staying the ordinary reflexes of the eye. In addition, there is another action called template matching. Such that, the pattern is compared with the other patterns saved in a database until either a matching pattern is discovered and the human being is recognized, or no match is located and human being still unknown. All those developments the performance for contrasting images by linearly-scale of time, and size. So, need to the error rates like False reject 156 http://www.i-jim.org Paper—Identification Based on Iris Detection Technique rate (FRR) thatcomputes the biometric measurement from live-failure to match the pat- terns, and the False accept rate (FAR) that computes the measurement from the live closes to another subject’s pattern which a right match is stated by mistake [28]. 2 Identification based iris detection using deep learning Iris as a biometric measure has many compensations over other measures. The iris is protected from different environmental situations. distinctiveness and unpredictabil- ity, and it has aconceptual biometric system. Deep learning is an efficient and powerful technique that is broadly employed in computer vision applications of account to the capability to obtain distinguishing characteristics from an eye-iris image in this work, the proposed method of iris identification based deep learning and iris recognition based CNN method are compared to get a decision of high performance and accuracy. 2.1 Datasets of iris image There is a sufficient quantity of database available to assist in assessing progress procedure and confirm heftiness. Most of these databases are obtained in the near-in- frared or visible spectrum. So, Images-databases gathered in no-compacted and com- pacted configures. Usually are caught in a variety of situations. Table 1 provides the catalog of not many of the eye image databases existing[29]. CASIA-V4 that contains 2576 NIRYESI-KEMB-100 Double-CAMERA extension to precursors but includes iris-images with various age stages 2,446 patterns of 142 subjects. Each sample was catching the higher division of a person-face and gripped together two eye-irises. Table 1. Some iris database Configure No. images No. topics Decree imaging device Name - 3200 200 640*480 LGIris Access Bio Sec .jepg 22,548 1614 640*480 OKI iris-passh CASIA-V3 .jepg 32,537 3284 640*480 IKEMB-100 CASIA-V4 .bmp 2240 224 320*240 JirisJPC1000 ITTD .Jpeg 3953 244 640*480 LG2000 ICE .bmp 289 1445 320*240 IrisAcess2200 MMU .tiff 11,101 522 400*300 CanonEOS5D UBIRIS-V2 .Jpeg/bmp 1540 79 2048*1360/1000*776 Canon EOS10D/ISG UTIRIS .bmp 220 945 640*480 Iris Pass WVU The patterns are obtained from 3 meters. All the right eye-images for each topic are composing the set of training, so the testing set includes each left eye-images [30]. Multimedia-University (MMU) dataset is collected from a publicly available database which is composed of iris patterns that can be used for training process based biometric identification models. This dataset comprises 460 images for 46 persons with 5 images for each of the left and right eyes[31]. Iris images are color with the size of 320×240 iJIM ‒ Vol. 16, No. 24, 2022 157 Paper—Identification Based on Iris Detection Technique pixels for each image and with depth as 3 dimensions or channels. IIT Delhi dataset consists of 1120 images and is prearranged as 224 specific covering, all images are transformed into numbers discriminating [32]. 2.2 Iris identification method using UniNet The methodology of implementing the model is divided into three stages as shown in Figure 1. All eye-iris images are subject to some pre-dealing out before the full con- volution network is began then the following identification stage. Fig. 1. Block diagram of identification stages Preprocessing iris images stage. Preprocessing stage is an actual essential role in the classification or detection development, so this stage comprises the following tasks: First: Iris region detection task. Duo methods are applied to find out the eye-iris region, one of these is to find out the region by window-schema. The pupil and iris are detected within a window. Stipulation this schema isn’t achieved to locate exact ending- results, another schema is employed, which deals with the complete eye-iris image as required-region i.e: the same steps but no dipping, there are many steps for this task, as shown in Figure 2. Detection Step of eye-pupil after eye-iris detection of eye-iris ap- plied Modified Hough Transform (MHT)[33], So, chosen value=50 as terminus to get binary eye-iris image subsequently estimated center and radius of a pupil by CH trans- form. Apply HT for detecting the outer iris. The amount of pupil radius values was 15– 50, but for iris was 55-100[34]. These steps are applied to different types of iris-images as shown in Figure 3. Stage1 Prepare Iris Images Stage1 Image Preprocessing stage2 Feature extraction St ag e3 Matching Stage4Identification Final Stage Get Result 158 http://www.i-jim.org Paper—Identification Based on Iris Detection Technique Fig. 2. Iris region detection steps Fig. 3. Iris detection results Second: Normalization task. As a part of preprocessing stage, the normalization task is employed to provide a permanent dimension vector of attributes that is considered as input-record to learning-phase after the iris region detection task was completed. Nor- malization is implemented by applying Daugman’s-rubber-sheet modeling that con- verts the data-iris image from cartesian to polar coordinates[33].Daugman’s mapping receipts all position (x, y) within detect-region to a pair of polar-coordinates (r, θ) with stabilized[35, 36].The converting process is represented as eq (1)(2)(3). iJIM ‒ Vol. 16, No. 24, 2022 159 Paper—Identification Based on Iris Detection Technique I�x(r, θ), y(r, θ)� → I(r, θ) (1) x(r, θ) = (1 − r)xp(θ)rxi(θ) (2) y(r, θ) = (1 − r)yp(θ)ryi(θ) (3) Where, r is in between [0, 1] 𝜃𝜃 is the angle in between [0, 2π] I(𝑥𝑥,𝑦𝑦) is the concentration rate at(𝑥𝑥,𝑦𝑦) The constraints 𝑥𝑥p, 𝑥𝑥l, 𝑦𝑦p, and 𝑦𝑦 l is organizing of the eye-pupil and eye-iris margins next to 𝜽𝜽 track as illustration Figure 4. Fig. 4. Normalization task results Feature extraction stage. UniNet is a modified deep learning structure that is used together masking of eyeiris-region and features mining, which is rooted in fully-con- volutional networks (FCN). Uni-Net structure is assembled of 2 sub-networks, a feature sub-network (Feat.Net) and a masking network (Mask.Net) as illustrated in Figure 5. Both are constructed on FCN. The major units of FC-Nets are multi layers of convolu- tional, pool layers unit, active functions unit, and the rest. Merging the up-sampling layers, fully-CNets is capable of performing all pixel’s prediction. Feat Net is intended for mining discriminative eye-iris attributes, that are utilized in the matching-stage. Mask Net is established to achieve no-eye-iris region masking for returned to normal eye-iris images, that observed the same as trouble for segmentation semantically ap- proach. The vital of masking grantees for loss function and matching-stage. The final- result of this stage is explained in Figure 6. 160 http://www.i-jim.org Paper—Identification Based on Iris Detection Technique Fig. 5. UniNet architecture [23] Fig. 6. Feature extraction stage Triplet network learning. A triplet network is executed to be trained the kernels within Feat.Net. Where three like Uninets, where weights are reserved equal through learning, are located in equivalence to ahead and back-propagate data with incline for anchor, both of them +ve with -ve samples correspondingly. Anchor+ve (AP)-pair must appear of matching someone though anchor-ve (AN)-pair appears of dissimilar some- one. Triplet-loss procedure challenges to lessen anchored distance meanwhile enlarge the anchor-ve distance. To guarantee extra proper with authentic control in making of eye-iris attributes-features based FC-Nets function, the triplet-loss is improved by inte- grating a bit-shifting process called (Extensive-Triplet-Loss (ETL))[16]. ETL process. Earliest loss work for networks of triplet describes using eq (4): 𝐿𝐿 = 1𝑁𝑁 ∑ ��𝑓𝑓𝑖𝑖 𝐴𝐴 − 𝑓𝑓𝑖𝑖 𝑃𝑃� 2 − �𝑓𝑓𝑖𝑖 𝐴𝐴 − 𝑓𝑓𝑖𝑖 𝑁𝑁� 2 + 𝛼𝛼� + 𝑁𝑁 𝑖𝑖=1 .. (4) Such N is no. of triplet patterns in a small set 𝑓𝑓𝑖𝑖 𝐴𝐴,𝑓𝑓𝑖𝑖 𝑁𝑁, and𝑓𝑓𝑖𝑖 𝑃𝑃are attributes plans +ve and –ve anchor in ith triplet correspondingly. [+] is equal to max (•, 0). α is a factor to manage preferred border of anchor+ve,anchor-ve distance. So, Improving the original loss leads to minimizing anchor+ve and maximizing an- chor-ve distance until their border is more of a specific value. By means of Euclidean- distance being variation metric is not as efficient as required. Applying spatial attrib- utes-features is proposed that have similar ruling with the record-enter; the matching procedure deals with no-eyeiris region masking and straight shifting, that are often pragmatic in eye-iris patterns. So, the earliest triplet-loss task is comprehensive, and cites as ETL. See eq (5), and Masked-Distance (MMSD) method, termed as eq (6): iJIM ‒ Vol. 16, No. 24, 2022 161 Paper—Identification Based on Iris Detection Technique 𝐸𝐸𝐸𝐸𝐿𝐿 = 1𝑁𝑁 ∑ [𝐷𝐷(𝑓𝑓𝑖𝑖 𝐴𝐴, 𝑓𝑓𝑖𝑖 𝑃𝑃) − 𝐷𝐷(𝑓𝑓𝑖𝑖 𝐴𝐴, 𝑓𝑓𝑖𝑖 𝑁𝑁) + 𝛼𝛼]+. . .𝑁𝑁𝑖𝑖=1 (5) Where D (𝑓𝑓1,𝑓𝑓2) is Minimum Shifted 𝐷𝐷(𝑓𝑓1, 𝑓𝑓2) = {𝐹𝐹𝐷𝐷(𝑓𝑓𝑏𝑏 1, 𝑓𝑓2)}−𝐵𝐵≤𝑏𝑏≤𝐵𝐵 𝑚𝑚𝑖𝑖𝑚𝑚 (6) A lot of arithmetical functions including derivations are applied to guarantee that main-features are trained in legal eye-iris zones, whereas no-eyeiris zones will do not take into account. after everything else the last convolutional-tier, a sole-channel attrib- ute-record is produced to gauge the matches of eye-iris images. Third: Matching task. To get the correct match of iris images, apply some steps as follows: 1. Binarized the iris-image features which are a real value to get less storage and the (0,1) values is a further confrontation to illumination change, distorting, and another underlying clatter. 2. Compute the Hamming-distance of the attributes-plan of step1 and comprehensive masks. See eq (7): 𝐷𝐷𝐻𝐻 = ∑ |(𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝐴𝐴⨁𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝐵𝐵)∩(𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝐴𝐴∩𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝐵𝐵)|𝐾𝐾𝑖𝑖=1 ∑ (𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝐴𝐴∩𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝐵𝐵)𝑘𝑘𝑖𝑖−1 (7) Where, 𝐷𝐷𝐻𝐻 =“0” for no-eyeiris-regions, and “1” for eyeiris-regions. Fourth: Discussion of identification task based matching task. The assessment if duo eye-irises are belonged to an exact person is ended by pre-identified threshold-rate. Based on the identical results got using NB classifier if the features are matched with the database, then the person is accepted, if not the person is rejected. 2.3 Iris recognition method using CNN [5] In this method, the learning based 2D-CNN model is classified images of eye-iris based eye features were extracted. Iris dataset is gathered, enhanced using histogram equalization (HE) schema with contrast limited adaptive equalization (CLAHE) sche- mas, and augmented. After that, CNN model classifies the iris images after extracted features. The construction of CNN consists of features-mining that includes convolu- tional-layers and ReLu-layers, factors-reduction includes pooling-layers, classifier of features-mining includes FC-layer and Softmax-layer. So, applying backpropagation of errors schema and adjusted moment estimation-Adam for achieving training phase and weights renovating. Iris recognition steps using CNN. Data augmentation is a very important technique for creating the better database and enlarging its size. This work handled dataset in- cludes 460 iris images then augmented images for maximizing the size of the dataset by creating adapted editions of images like images cropping-task, shifting-task, scaling- task, shearing-task, and flipping-task. After that adjust the contrast in the original image using, HE and CLAHE techniques. Then the total number of the dataset became 4600 when completing the image preprocessing and augmentation. The CNN structure has ten layers, beginning with the input layer and continuing with two convolution layers, 162 http://www.i-jim.org Paper—Identification Based on Iris Detection Technique two activation RuLe, duo pool-layers, one FC- layer, SoftMax layer, and finally the output-result-layer, as demonstrated in Figure 7. Then back-propagation algorithm is used for the training procedure and Adam optimizer for updating the weight. The total set of data was split into 0.80 of all images to do train mode with 0.20 of all images for testing-set. The images were trained using 4 as bunch size, learning value-rate of 0.0001, and utmost of 400 epochs using Adam optimizer. Fig.7.CNN-layers diagram for iris recognition [5] 3 Results analysis This work aims to achieve high accuracy with compete speed, therefore introducing two methods of iris detection that depend on deep structured learning to compare the efficient performance of these two methods. So, in this method, has been hired with a lot of preprocessing methods to get more accurate and also faster than the state of using the image without preprocessing. Also using two methods for pupil detection made this layer1 •Image input layer2 •Conv layer3 •Relu layer4 •Maxpooling layer5 •Conv layer6l •Relu layer7 •Maxpooling layer8 •Fully Connected(fc) layer9 •Softmax layer10 •Classoutput iJIM ‒ Vol. 16, No. 24, 2022 163 Paper—Identification Based on Iris Detection Technique work more accurate. In addition, the use of the normalization process by removing di- mensional in consistency had a role to speed up and accurate results. At a matching stage, the detection task uses (0,1) as attribute-features don’t be low in the achievement contrasted with actual-valued of attributes to even yield enhancements in some standard dataset scenarios. This method was applied to various datasets with images caught by a particular camera-device. The Testing set of CASIA-V4 produces 20,702 valid matches up and 2,969,533 cheat matches up. Testing set IIT-D contains 2,240 valid matches up and 624,400 cheat match up. Cross-DB (i.e training is done by dataset dif- fers from the tested one) and Within-DB arrangements are used for testing this work, The goal of Cross-DB is to check the performance of the framework if limited training samples are existing. Within Database using only one set of data to train the net then applying fine-tune it based differ training sets. Then fine-tuned model is assessment on the special testing set. The fine-tuned models of within-DB arrangement are done well rather than cross-DB, with expected higher correlation of image quality between the training and testing sets. All experimental outcomes are generated as many-to-many similar policies. The dextractor of attributes was achieved high accuracy also capable of generality. The final accurate-results of this work structure are measure up to other related works that aim to its state of the art method with good performance. Assessment includes SIRIS [37],Gabor filter (GABOR) [38, 39], and outcome of Cross and within a dataset of this method .The output of that method is evaluated against other related works that aim to its state-of-the-art method that is the high ranking of standards. The accuracy rates[40] is regard to the rate of recognition (FAR, EER, FRR),The results of the assessment are shown in Table 2. Table 2. Results comparison MMU CASIA IITD FRR% EER% FRR% EER% FRR% EER% OSIRIS 17.83 3.43 19.93 6.39 1.6 1.11 GABOR 18.58 5.47 20.72 7.71 1.81 1.38 CROSS 14.37 3.12 13.22 5.03 0.80 0.61 WITN 9.92 5.27 11.9 4.74 1.16 0.7 Table 2 illustrates the FRR is the chance that is false to an exact person and theEqual Error Rate (EER) which is computed by an all-against-all comparison: As a result, as the ratio of missing data is increased, more and more false rejections of valid data, resulting in increased EER. The percentage values of overall training accuracy were 96.5, and 98.1 for CROSS DB, and WITHN DB respectively. In the second method, increasing the number of epochs will improve the training process by achieving the best results and increasing the reliability but would be also increased the amount elapsed. The percentage values of overall training accuracy were 95.31% and a spending time 17.59 minutes for 400 epochs. Training and validation steps show that the overall ac- curacy is very good. Table 3 shows the FAR criteria for the different terminus of method on CASIA-V4, ITTD, and MMU dataset. 164 http://www.i-jim.org Paper—Identification Based on Iris Detection Technique Table 3. FAR measure with different terminus value in the method FAR Terminus(threshold) MMU ITT-D CASIA 0.71 0.64 10.61 CROSS 40 0.60 0.52 8.54 WTHIN 0.70 0.65 11.76 CROSS 50 0.61 0.54 10.43 WTHIN 0.70 0.65 12.8 CROSS 60 Table 4 illustrates the results of EER and accuracy after applying the steps of the method on the various eye images dataset. Table 4. Results of EER and accuracy AUC. EER 98.49% 4.93% CASIA-V4 99.12% 0.84% ITT-D 98.02% 3.43% MMU Table 5 illustrates the value accuracy[41, 42] rate of identification after applying the two mentioned methods to MMU dataset. Table 5. Accuracy-rate of identification for MMU dataset Accuracy Time consumption(sec) UniNet Method 98.845% 31.646 Identification basedCNN[5] 96.313% 11.576 4 Conclusion Due to the employment of couple of subnets for feature extraction and a mask for no-eyeiris parts, the usage of UNINET proves successful in eye-iris recognition. By formative and discarding no-eyeiris parts before the identical part, the matching(iden- tical) result is more fast and more truthful the effective pre-dealing out methods used to deliver UNINET input are another factor in these encouraging outcomes. Thestrategy yields good accuracy, although more preparation and investigation datasets are re- quired. Additionally, a high-declaration dataset was required. 5 Acknowledgment The authors would like to thank the University of Technology – Iraq www.uotech- nology.edu.iq and Mustansiriyah University – Iraq www.uomustansiriyah.edu.iq for the present work. iJIM ‒ Vol. 16, No. 24, 2022 165 Paper—Identification Based on Iris Detection Technique 6 References [1] E. Severo et al., "A benchmark for iris location and a deep learning detector evaluation," in 2018 international joint conference on neural networks (IJCNN), 2018: IEEE, pp. 1-7. https://doi.org/10.1109/IJCNN.2018.8489638 [2] D. Menotti et al., "Deep representations for iris, face, and fingerprint spoofing detection," IEEE Transactions on Information Forensics and Security, vol. 10, no. 4, pp. 864-879, 2015. https://doi.org/10.1109/TIFS.2015.2398817 [3] I. Buciu and A. Gacsadi, "Biometrics systems and technologies: a survey," International Journal of Computers Communications & Control, vol. 11, no. 3, pp. 315-330, 2016. https://doi.org/10.15837/ijccc.2016.3.2556 [4] N. Ortiz, R. D. Hernández, R. Jimenez, M. Mauledeoux, and O. Avilés, "Survey of biometric pattern recognition via machine learning techniques," Contemporary Engineering Sciences, vol. 11, no. 34, pp. 1677-1694, 2018. https://doi.org/10.12988/ces.2018.84166 [5] B. K. O. C. Alwawi and A. F. Y. Althabhawee, "Towards more accurate and efficient human iris recognition model using deep learning technology," TELKOMNIKA (Telecommunication Computing Electronics and Control), vol. 20, no. 4, pp. 817-824, 2022. https://doi.org/10. 12928/telkomnika.v20i4.23759 [6] R. Alinia Lat, S. Danishvar, H. Heravi, and M. Danishvar, "Boosting Iris Recognition by Margin-Based Loss Functions," Algorithms, vol. 15, no. 4, p. 118, 2022. https://doi.org/10. 3390/a15040118 [7] K. Sharma and R. Singh, "An Optimized and Robust Iris Recognition Algorithm for Biometric Authentication Systems," Journal of Scientific and Technical Advancements, vol. 4, no. 4, pp. 1-8, 2018. [8] N. Alseelawi, and H. T. Hazim, "A Novel Method of Multimodal Medical Image Fusion Based on Hybrid Approach of NSCT and DTCWT," iJOE, vol. 18, no. 03, p. 115, 2022. https://doi.org/10.3991/ijoe.v18i03.28011 [9] A. K. Jain, P. Flynn, and A. A. Ross, Handbook of biometrics. Springer Science & Business Media, 2007. https://doi.org/10.1007/978-0-387-71041-9 [10] C. M. Patil, "An Efficient Iris Recognition System to Enhance Security Environment for Personal Identification," vol., no, 2013. https://doi.org/10.20894/IJCNES.103.002.001.003 [11] S. H. Abbood, M. S. Rahim, and A. M.Alaidi, "DR-LL Gan: Diabetic Retinopathy lesions synthesis using Generative Adversarial Network," International journal of online and biomedical engineering, vol. 18, no. 3, pp. 151-163, 2022. https://doi.org/10.3991/ijoe.v18i03. 28005 [12] K. Wang and A. Kumar, "Toward more accurate iris recognition using dilated residual features," IEEE Transactions on Information Forensics and Security, vol. 14, no. 12, pp. 3233-3245, 2019. https://doi.org/10.1109/TIFS.2019.2913234 [13] I. A. Aljazaery, J. S. Qateef, A. H. M. Alaidi, and R. a. M. Al_airaji, "Face Patterns Analysis and Recognition System Based on Quantum Neural Network QNN," International Journal of Interactive Mobile Technologies, vol. 16, no. 8, 2022. https://doi.org/10.3991/ijim. v16i08.30107 [14] Z. Zhao and A. Kumar, "Towards more accurate iris recognition using deeply learned spatially corresponding features," in Proceedings of the IEEE international conference on computer vision, 2017, pp. 3809-3818. https://doi.org/10.1109/ICCV.2017.411 166 http://www.i-jim.org https://doi.org/10.1109/IJCNN.2018.8489638 https://doi.org/10.1109/TIFS.2015.2398817 https://doi.org/10.15837/ijccc.2016.3.2556 https://doi.org/10.12988/ces.2018.84166 https://doi.org/10.12928/telkomnika.v20i4.23759 https://doi.org/10.12928/telkomnika.v20i4.23759 https://doi.org/10.3390/a15040118 https://doi.org/10.3390/a15040118 https://doi.org/10.3991/ijoe.v18i03.28011 https://doi.org/10.1007/978-0-387-71041-9 https://doi.org/10.20894/IJCNES.103.002.001.003 https://doi.org/10.3991/ijoe.v18i03.28005 https://doi.org/10.3991/ijoe.v18i03.28005 https://doi.org/10.1109/TIFS.2019.2913234 https://doi.org/10.3991/ijim.v16i08.30107 https://doi.org/10.3991/ijim.v16i08.30107 https://doi.org/10.1109/ICCV.2017.411 Paper—Identification Based on Iris Detection Technique [15] A. S. Al-Waisy, R. Qahwaji, S. Ipson, S. Al-Fahdawi, and T. A. Nagem, "A multi-biometric iris recognition system based on a deep learning approach," Pattern Analysis and Applications, vol. 21, no. 3, pp. 783-802, 2018. https://doi.org/10.1007/s10044-017-0656-1 [16] R. Shi, Y. Wang, Y. Chen, and C. Wu, "An Efficient Recognition Method for Incomplete Iris Image Based on CNN Model," in 2018 IEEE SmartWorld, Ubiquitous Intelligence & Computing, Advanced & Trusted Computing, Scalable Computing & Communications, Cloud & Big Data Computing, Internet of People and Smart City Innovation (SmartWorld/SCALCOM/UIC/ATC/CBDCom/IOP/SCI), 2018: IEEE, pp. 1154-1159. https://doi.org/10.1109/SmartWorld.2018.00200 [17] M. Liu, Z. Zhou, P. Shang, and D. Xu, "Fuzzified image enhancement for deep learning in iris recognition," IEEE Transactions on Fuzzy Systems, vol. 28, no. 1, pp. 92-99, 2019. https://doi.org/10.1109/TFUZZ.2019.2912576 [18] M. Al-dabag, and R. Al-Nima, "Anticipating Atrial Fibrillation Signal Using Efficient Algorithm," International Journal of Online and Biomedical Engineering (iJOE), vol. 17, no. 2, pp. 106-120, 2021. https://doi.org/10.3991/ijoe.v17i02.19183 [19] M. Omran and E. N. AlShemmary, "An iris recognition system using deep convolutional neural network," in Journal of Physics: Conference Series, 2020, vol. 1530, no. 1: IOP Publishing, p. 012159. https://doi.org/10.1088/1742-6596/1530/1/012159 [20] A. Ignat and I. Păvăloi, "Experiments on iris recognition using SURF descriptors, texture and a repetitive method," Procedia Computer Science, vol. 176, pp. 175-184, 2020. https://doi.org/10.1016/j.procs.2020.08.019 [21] H. T. Hazim, and H. Salim "Secure Chaos of 5G Wireless Communication System Based on IOT Applications," International Journal of Online and Biomedical Engineering(iJOE), vol. 18, no. 12, pp. 89-102, 2022. https://doi.org/10.3991/ijoe.v18i12.33817 [22] V. Conti, L. Rundo, C. Militello, V. M. Salerno, S. Vitabile, and S. M. Siniscalchi, "A multimodal retina‐iris biometric system using the Levenshtein distance for spatial feature comparison," IET Biometrics, vol. 10, no. 1, pp. 44-64, 2021. https://doi.org/10.1049/bme2. 12001 [23] S. K. Singla and P. Sethi, "Challenges at different stages of an iris based biometric system," Songklanakarin Journal of Science & Technology, vol. 34, no. 2, 2012. [24] A. Al-zubidi, R. K. Hasoun, and S. Hashim, "Mobile Application to Detect Covid-19 pan- demic by using Classification Techniques: Proposed System," International Journal of Interactive Mobile Technologies, vol. 15, no. 16, pp. 34-51, 2021. https://doi.org/ 10.3991/ijim.v15i16.24195 [25] B. R. B. Navneet, "Challenges and Achievements in Iris Recognition: A General Study," NTERNATIONAL JOURNAL OF INNOVATIVE RESEARCH IN TECHNOLOGY, vol. 8, no. 11. [26] D. M. Rankin, B. W. Scotney, P. J. Morrow, and B. K. Pierscionek, "Iris recognition failure over time: The effects of texture," Pattern Recognition, vol. 45, no. 1, pp. 145-150, 2012. https://doi.org/10.1016/j.patcog.2011.07.019 [27] H. T. H. H. Alrikabi, "Enhanced Data Security of Communication System using Combined Encryption and Steganography," International Journal of Interactive Mobile Technologies, vol. 15, no. 16, pp. 144-157, 2021. https://doi.org/10.3991/ijim.v15i16.24557 [28] Eyetrackingupdate, "Beware of problems with iris recognition,"http://eyetracking update. com/2010/11/03/beware-problems-iris-recogniti on/ November 7, 2011. [29] J. J. Winston and D. J. Hemanth, "A comprehensive review on iris image-based biometric system," Soft Computing, vol. 23, no. 19, pp. 9361-9384, 2019. https://doi.org/10. 1007/s00500-018-3497-y [30] w. b. i. o. d. d. i. CASIAIris DatabaseV4. iJIM ‒ Vol. 16, No. 24, 2022 167 https://doi.org/10.1007/s10044-017-0656-1 https://doi.org/10.1109/SmartWorld.2018.00200 https://doi.org/10.1109/TFUZZ.2019.2912576 https://doi.org/10.3991/ijoe.v17i02.19183 https://doi.org/10.1088/1742-6596/1530/1/012159 https://doi.org/10.1016/j.procs.2020.08.019 https://doi.org/10.3991/ijoe.v18i12.33817 https://doi.org/10.1049/bme2.12001 https://doi.org/10.1049/bme2.12001 https://doi.org/10.3991/ijim.v15i16.24195 https://doi.org/10.3991/ijim.v15i16.24195 https://doi.org/10.1016/j.patcog.2011.07.019 https://doi.org/10.3991/ijim.v15i16.24557 https://doi.org/10.1007/s00500-018-3497-y https://doi.org/10.1007/s00500-018-3497-y Paper—Identification Based on Iris Detection Technique [31] w. k. c. d. n. m.-i.-d. MMUirisdataset. [32] w. c. p. e. h. c. I. D. I. h. IIT Delhi IrisDatabase. [33] K. N. Jassim and A. K. Nseaf, "PERFORMANCE AND RELIABILITY FOR IRIS RECOGNITION BASED ON CORRECT SEGMENTATION," Journal of Theoretical & Applied Information Technology, vol. 93, no. 2, 2016. [34] A. J. K.KUDER, K. NASSIF , A. KHUDHUR, M. OUDELHA, "DNN FOR IRIS RECOGNITION SYSTEM BASED ON VIDEO: SSAE AND BI-PROPAGATION NN MODELS," Journal of Theoretical and Applied Information Technology, vol. 93, no. 2, 2016. [35] J. G. Daugman, "High confidence visual recognition of persons by a test of statistical independence," IEEE transactions on pattern analysis and machine intelligence, vol. 15, no. 11, pp. 1148-1161, 1993. https://doi.org/10.1109/34.244676 [36] Y. Chen, C. Wu, and Y. Wang, "T-center: a novel feature extraction approach towards large- scale iris recognition," IEEE Access, vol. 8, pp. 32365-32375, 2020. https://doi.org/10.1109/ ACCESS.2020.2973433 [37] N. Othman, B. Dorizzi, and S. Garcia-Salicetti, "OSIRIS: An open source iris recognition software," Pattern Recognition Letters, vol. 82, pp. 124-131, 2016. https://doi.org/10.1016/ j.patrec.2015.09.002 [38] J. Daugman, "How iris recognition works," in The essential guide to image processing: Elsevier, 2009, pp. 715-739. https://doi.org/10.1016/B978-0-12-374457-9.00025-1 [39] S. H. Shaker and F. Q. Al-Khalidi, "Human Gender and Age Detection Based on Attributes of Face," International Journal of Interactive Mobile Technologies, vol. 16, no. 10, 2022. https://doi.org/10.3991/ijim.v16i10.30051 [40] R. Ogla, A. A. Saeid, and S. H. Shaker, "Technique for recognizing faces using a hybrid of moments and a local binary pattern histogram," International Journal of Electrical and Computer Engineering, vol. 12, no. 3, p. 2571, 2022. https://doi.org/10.11591/ijece.v12i3. pp2571-2581 [41] S. H. Shaker, Al-Khalidi Farah, Q., Mahdi, A., "DIAGNOSIS OF DIABETES MELLITUS BASED COMBINED OF FEATURE SELECTION METHODS," Journal of Theoretical and Applied Information Technology, vol. 100, no. 13, 2022. [42] M. Kandee, P. Boonbrahm, and V. Tantayotai, "Development of Virtual Pulse Simulation for Pulse Diagnosis Studies," International Journal of Interactive Mobile Technologies, vol. 12, no. 7, 2018. https://doi.org/10.3991/ijim.v12i7.9640 7 Authors Dr. Shaimaa Hameed Shaker, computer sciences-pattern recognition-2006- Uni- versity of technology – Iraq Baghdad. Received the BSC in computer science by uni- versity of technology – Iraq in 1991. She graduated with the Master of Science in com- puter science-visual cryptography in 1996 by University of technology – Iraq and fin- ished the Ph.D. degree in Computer sciences – pattern recognition 2006, respectively. Her interest in pattern recognition, cryptography and data security, information hiding, image processing, and bioinformatics. Currently, she is ahead of the network manage- ment department of computer science college – University of technology, she is a lec- turer. Can be contacted at email: Shaimaa.h.shaker@uotechnology.edu.iq. 168 http://www.i-jim.org https://doi.org/10.1109/34.244676 https://doi.org/10.1109/ACCESS.2020.2973433 https://doi.org/10.1109/ACCESS.2020.2973433 https://doi.org/10.1016/j.patrec.2015.09.002 https://doi.org/10.1016/j.patrec.2015.09.002 https://doi.org/10.1016/B978-0-12-374457-9.00025-1 https://doi.org/10.3991/ijim.v16i10.30051 https://doi.org/10.11591/ijece.v12i3.pp2571-2581 https://doi.org/10.11591/ijece.v12i3.pp2571-2581 https://doi.org/10.3991/ijim.v12i7.9640 Paper—Identification Based on Iris Detection Technique Dr. Farah Qais Al-Khalidi, computer sciences – image processing-2012- Univer- sity of technology Iraq-Baghdad. Mustansiriyah University, from January 2005 – Pre- sent, she is in the computer science Department as Assistance prof. in computer science, she received PHD. From Sheffield hallam University – January 2007 – January 2012. Field of study is information technology (computer science). Can be contacted at email: farahqaa@uomustansiriyah.edu.iq. Dr. Raheem Ogla has been a faculty member at the Department of Computer Sci- ence since 2014. He graduated from the University of Technology, Baghdad, Iraq. He got his master’s degree in Computer Science in 2004. His MSc. thesis was in “Design and Implementation of Fire Artillery Software System”. In 2010, He received Degree his Doctorate in Philosophy of Computer Science at the University of Technology (UOT), Baghdad, Iraq. The Ph.D. My dissertation was in “Digital Video Compression Scheme Using Enhanced Fractal Coding”. He can be contacted at email: Ra- heem.A.Ogla@uotechnology.edu.iq. Article submitted 2022-09-23. Resubmitted 2022-11-05. Final acceptance 2022-11-06. Final version pub- lished as submitted by the authors. iJIM ‒ Vol. 16, No. 24, 2022 169 mailto:farahqaa@uomustansiriyah.edu.iq