MISSING CHILD IDENTIFICATION SYSTEM USING DEEP LEARNING AND MULTICLASS SVM

Abstract

In India alone, hundreds of children disappear every year, and many of them are never traced.
This paper introduces a facial recognition system based on deep learning to help trace missing
children. The general public can post pictures of suspected missing children on a common
webpage, along with comments and location information. The system then compares the posted
picture with registered missing child images in the database. A deep learning model, which has
been trained on publicly available facial images, is employed for precise identification. The
method employs Convolutional Neural Networks (CNN), namely the VGG-Face deep
architecture, to extract facial features. Unlike other deep learning models, this system uses CNNs
as high-level feature extractors and a trained Support Vector Machine (SVM) classifier for child
recognition. The chosen VGG-Face model provides strong identification, working well under
changes in noise, lighting, contrast, occlusion, facial positions, and aging. The new approach
outperforms existing face recognition-based methods, with a startling 99.41% classification
accuracy. It has been applied to 43 cases of child identification and proved it can significantly
enhance missing child recovery efforts. By employing this system, the authorities are better
equipped to trace missing children and reunite them with their families, offering a major
advancement in child safety and protection.