1139 Implementation of Preprocessing In Face Detection and Crime Avoidance Using Deep Learning Techniques

Abstract

Identification and recognition of criminals is a tough and time-consuming process at any crime scene. Criminals can be recognized through fingerprints, DNA evidence, captured images from CCTV or other surveillance equipment, or eyewitness testimonies. However, methods like fingerprint matching, DNA analysis, and face recognition from images require an already established and reliable database for effective identification. It is not just criminal investigations that rely on such technologies; various human recognition systems for access control, attendance tracking, and security verifications also require a strong database and image capturing systems. In this article, a method for recognizing human faces along with estimating age and gender using facial features is presented. Since images are multidimensional and may be influenced by several external factors such as lighting, angle, and noise, developing an accurate and reliable recognition model remains a challenging task. To enhance the system’s accuracy, various preprocessing techniques and feature extraction methods are employed to convert the input image into pixel-level representations suitable for machine learning models. The processed data is then fed into a Convolutional Neural Network (CNN) for recognition and classification based on age and gender. The primary goal of preprocessing is to eliminate redundancy, reduce distortions, and improve the quality of the images. Additionally, this work analyzes the impact of three different preprocessing techniques on the overall image classification performance, providing insights into their effectiveness in enhancing recognition accuracy.