Machine Vision System for Detection of Guns and Knives in Criminal Acts

León León R.A.; Polo Riveros J.J.; Haro Reyes J.C. · Memorias de la Conferencia Iberoamericana de Complejidad, Informatica y Cibernetica, CICIC (2025) · DOI: 10.54808/cicic2025.01.252

Resumen

This study presents the development of a computer vision system designed to detect weapons, such as pistols and knives, in real time, aiming to enhance public safety. Using the YOLOv8 model, the system integrates advanced image processing techniques and convolutional neural networks, leveraging Roboflow for image labeling and Google Colab for model training. Initial tests showed an 84.6% effectiveness in identifying weapons in video, despite challenges related to lighting variability and camera resolution. Nevertheless, the system shows significant potential to improve security in urban environments. It is concluded that upgrading hardware quality and expanding the dataset could make the model even more robust and reliable for real-world scenarios. Copyright 2025. © por el International Institute of Informatics and Systemics.

Instituciones: Universidad Privada del Norte