Object Detection and Movement Patterns Using Neural Networks and Genetic Algorithms for the Identification of Armed Robbery

  • Walter Leturia-Rodriguez Universidad Privada Antenor Orrego
  • Luis Urrelo-Huiman Universidad Privada Antenor Orrego
Keywords: Machine Learning, Object Detection, Movement Pattern Detection


In Latin America there are 42 of the 50 most violent cities in the world, in Peru in 2019 9.7% of criminal acts with firearms were carried out in urban areas and in cities with 20,000 or more inhabitants the percentage rose to 10.2%, but the complaints, due to lack of evidence, generated in cities like Lima, only 19.46% of arrests. The members of the police force have devices, vehicles and tools that allow them to carry out their functions in a safe manner, however, they do not have an effective mechanism, which allows to identify an armed robbery and to concentrate their efforts on carrying out a timely intervention. Therefore, the present research develops an algorithm based on Recurrent Neural Networks with OpenCv / YOLOv3 combined with the Genetic Algorithms technique for the detection of objects and movement patterns with 96.5% accuracy, allowing early detection of a crime perpetrated under the modality of armed robbery.


Download data is not yet available.
Research Articles for the Regular Issue