Int J Performability Eng ›› 2021, Vol. 17 ›› Issue (4): 379-393.doi: 10.23940/ijpe.21.04.p6.379393

• Original article • Previous Articles     Next Articles

Intelligent Optimization of Latent Fingerprint Image Segmentation using Stacked Convolutional Autoencoder

Chhabra Meghaa,*(), Shukla Manoj Kumarb, and Ravulakolluc Kiran Kumarc   

  1. a AIIT, Amity University, Noida, 201301, India
    b ASET, Amity University, Noida, 201301, India
    c School of Computer Science, University of Petroleum and Energy Studies, Dehradun, 248007, India
  • Contact: Chhabra Megha


In recent times, image forensics have contributed to multiple research areas like latent fingerprint forensics. Latent fingerprint forensics is performed by law enforcement agencies. Latent fingerprint matching and identification of criminals depends highly on accurate detection and therefore segmentation of fingermarks. A novel algorithm is designed a) for efficient pre-feature extraction as the early-distinction of the structure of interest using saliency and color-map based information, thereby channeling lesser data for patch-based classification, b) for reliable post-feature extraction-based classification and segmentation task via a dropout-based regularized stack of convolutional autoencoder, which helps in automating optimal feature selection and reliable object detection with reduced overfitting. An intelligent parameter standardization of the proposed technique that improves the efficiency and effectiveness of the segmentation process is proposed to demonstrate the repeatability of the system using cross-validation. The experimentation is performed on the IIIT-D CLF database. By addressing the need for intelligent distinction of the structure of interest in latent fingerprint and impact of patch size in the pre-feature extraction phase along with optimal feature selection and object detection using a stacked convolutional autoencoder, the proposed work improved the latent fingerprint segmentation and detection technique.

Key words: latent fingerprint, segmentation, convolutional neural network, autoencoder, image enhancement, image classification.