Design and Performance Analysis of Latent Fingerprint Segmentation Algorithms

Abstract

Latent fingerprints are the necessary evidence utilized by law enforcement agencies for identifying suspects LFP segmentation includes marking out the entire foreground region preciously in an LFP image Because of the poor quality of images and complexity in the background LFP segmentation is one of the most complex processes in an automatic LFP recognition system Hence two approaches are performed in this research one method for enhancement and another method for segmentation of LFP The first work involves LFP enhancement based on Bent Identity Convolutional Neural Network with Spatial Pyramid Max Pooling This procedure involves the integration of Region of Interest estimation Anisotropic Gaussian Filter based Pre filtering Fingerprint alignment using Sobel Filter Intrinsic Feature patch extraction using BI CNN Graph Attention network based Similarity Estimation followed by image reconstruction and feedback module The implementation tool used in this work is the PYTHON The proposed optimized BI CNN framework tested on dual public datasets namely IIITD latent fingerprint and IIITD MOLF which have shown enhanced outcomes The IIITD latent fingerprint database obtained 83 point 33 percent on Rank 10 accuracy and the MOLF database obtains 39 point 33 percent on Rank 25 accuracy Second phase of work involves segmentation Here the ridges of the latent fingerprint are segmented from the noisy background by defining various features such as saliency feature quality feature image intensity feature ridge feature and gradient feature These features are extracted and selected by utilizing the adaptive equalizer optimization technique The deep attention based bidirectional long short term memory network is used for segmentation having world cup optimization for weight update The designed model is implemented in Python platform that have yield the segmentation accuracy of 98 point 89 percent and having very low false detection and missed detection ratio which is better than the other extant methods newline