Anomalies Identification in Images from Security Video Cameras Using Mask R-CNN

In this work we developed a system to identify anomalies in images from video security cameras in an urban environment. Initially people are detected in the images using Mask R-CNN. From the binary mask are extracted characteristics of the people so that the anomalies can be detected. In order to fa...

ver descrição completa

Detalhes bibliográficos
Autores: Minari, G., Silva, F., Pereira, D., Almeida, L., Pazoti, M., Artero, A. [UNESP], Albuquerque, V de
Formato: artículo
Estado:Versión publicada
Fecha de publicación:2020
País:Brasil
Recursos:Universidade Estadual Paulista (UNESP)
Repositorio:Repositório Institucional da UNESP
Idioma:inglés
OAI Identifier:oai:repositorio.unesp.br:11449/195361
Acesso em linha:http://dx.doi.org/10.1109/TLA.2020.9082724
http://hdl.handle.net/11449/195361
Access Level:acceso abierto
Palavra-chave:Mask R-CNN
CNN
HOG
People characteristics extraction
Intrusion detection
Facial recognition
Descrição
Resumo:In this work we developed a system to identify anomalies in images from video security cameras in an urban environment. Initially people are detected in the images using Mask R-CNN. From the binary mask are extracted characteristics of the people so that the anomalies can be detected. In order to facial recognition we used Facial Landmarks so that the system knows the residents and authorized people avoiding the false anomalies. We considered four anomalies in this work: the act of jumping a wall, standing for a long time in front of the residence, walking thru the sidewalk several times and entering a place without permission.