INTEGRATION DE LA RECONNAISSANCE FACIALE 2D/3D ET DE LA GEOLOCALISATION DANS UNE ARCHITECTURE IOT POUR L’IDENTIFICATION AUTOMATIQUE DES AUTEURS D’INFRACTIONS ROUTIERES

Increase in road traffic violations in cities is a serious problem for the public authorities, especially in developing countries where systems of surveillance are deficient. This work presents an intelligent system that offers automated enforcement of road violations by integrating 2D/3D face recognition with geofence and distributed IoT. The objective of this approach is to enhance precision in identifying infringements, spatio-temporal traceability, and automation of infringement handling. According to this architecture, machine learning models are applied at two levels of work, i.e., 2D/3D facial recognition level for the purpose of biometric feature extraction and decision-making level for intelligent fusion between biometric information with geospatial and kinematic data; while lightweight models operate on edge side in order to improve data preprocessing. The results demonstrate a significant improvement in identification accuracy and processing time. The system represents an innovative and reproducible solution for intelligent road infrastructures. The methodology includes an in-depth literature review, theoretical modeling, implementation of a prototype based on embedded modules (ESP32-Cam, GPS) and a FastAPI backend, as well as real-world testing in the city of Kananga.