Criminal Identification using Facial Recognition

The large volume and often poor quality of CCTV footage pose significant challenges for criminal identification, making manual review impractical and leading to missed opportunities. To address this, we propose a system that enhances and analyzes keyframes from CCTV feeds using AI and machine learning. Frame differencing is used to extract candidate frames, which undergo preprocessing and enhancement using ESRGAN to improve image quality and detail. The enhanced keyframes are stored in a PostgreSQL database, where facial embeddings are generated using recognition and mapping algorithms. Identification is performed through vector similarity analysis, enabling fast and accurate suspect detection. This system significantly improves image quality, criminal recognition, and identification accuracy, bolstering law enforcement’s investigative capabilities while promoting public safety and crime prevention.