FKIMNet: A Finger Dorsal Image Matching Network Comparing Component (Major, Minor and Nail) Matching with Holistic (Finger Dorsal) Matching

Abstract

Current finger knuckle image recognition systems, often require users to place fingers' major or minor joints flatly towards the capturing sensor. To extend these systems for user non-intrusive application scenarios, such as consumer electronics, forensic, defence etc, we suggest matching the full dorsal fingers, rather than the major/ minor region of interest (ROI) alone. In particular, this paper makes a comprehensive study on the comparisons between full finger and fusion of finger ROI's for finger knuckle image recognition. These experiments suggest that using full-finger, provides a more elegant solution. Addressing the finger matching problem, we propose a CNN (convolutional neural network) which creates a 128-D feature embedding of an image. It is trained via. triplet loss function, which enforces the L2 distance between the embeddings of the same subject to be approaching zero, whereas the distance between any 2 embeddings of different subjects to be at least a margin. For precise training of the network, we use dynamic adaptive margin, data augmentation, and hard negative mining. In distinguished experiments, the individual performance of finger, as well as weighted sum score level fusion of major knuckle, minor knuckle, and nail modalities have been computed, justifying our assumption to consider full finger as biometrics instead of its counterparts. The proposed method is evaluated using two publicly available finger knuckle image datasets i.e., PolyU FKP dataset and PolyU Contactless FKI Datasets.

Deep rule-based classifier for finger knuckle pattern recognition system

“…11 Curve ROC for all sum score level fusion achieved highest recognition rate (rank-1) compared to scheme designed in Zeinali et al (2014). In addition, the presented system in this study achieved 99.65% of rank-1, which is better than FKIMNet proposed in Thapar et al (2019). It can also be noted that work in Chlaoua et al (2019) achieved the highest recognition rate 100% for all fingers, but if we observe for two fingers (e.g., LIF-RIF or LIF-LMF) the proposed system reached a good performance.…”

“…We can notice in Table 9 that the proposed system under verification mode attained lowest EER that is 0.19%. compared to EER by systems in Jaswal et al (2017b) and Thapar et al (2019) for all fingers. In identification mode, for the modalities (LIF-RIF), (LIF-LMF) and all finger the proposed system Fig.…”

“…The data augmentation techniques of random histogram equalization and dropout layer have been implemented to prevent over fitting during training of the proposed scheme. Thapar et al (2019) designed a scheme based on deep learning named FKIMNet for Finger Dorsal Image Matching Network. Chlaoua et al (2019) have investigated a simple deep learning method named principal component analysis Network (PCANet) with SVM classifier.…”

PixISegNet: pixel‐level iris segmentation network using convolutional encoder–decoder with stacked hourglass bottleneck

“…In future work, we would like to create an end‐to‐end deep learning iris‐based recognition system. Such a system would consist of three modules, (i) PixISegNet for pixel‐level segmentation (occlusion exclusion), (ii) an encoder–decoder model for quality assessment and enhancement of the images [50] and finally (iii) a Siamese‐based matching network [49]. Since the individual modules not only learn to optimise their task but also learn to provide an output that optimises the final recognition problem, making our network more generalised.…”

Selection of optimized features for fusion of palm print and finger knuckle‐based person authentication

“…Nowadays, contactless acquisition‐based palm or knuckle recognition algorithms are quite challenging, which can improve their user‐friendliness and scope in a variety of security applications. Also, the efforts have been made on cross spectral image matching, liveness detection, and employing deep convolution neural network‐based end to end biometric systems to improve the matching performance (Jaswal, Nath, et al, 2017; Thapar, Jaswal, & Nigam, 2019).…”