Seguimiento y evaluación de personas en ambientes cerrados / abiertos Thesis

short description

  • Undergraduate thesis

Thesis author

  • Cruz Angel, Bryan Santiago

abstract

  • Currently, in a constantly evolving world, we have witnessed the impact and growth of technology. What is known as the technological revolution has led to the creation of online payment systems, Artificial Intelligence (AI), BigData, the internet of things, among others. These tools facilitate and optimize processes that normally generate more economic, environmental and social expenses, especially the use of AI technologies presents a great performance in the detection of objects through images, when it comes to tracking people, the general norm dictates that it must be carried out manually, where a worker is in charge of keeping a control of the surveillance cameras, through which a subject can be identified, their position, and their physical condition, this control is effective but tedious, in The focus of this directed work seeks to automate this work, especially in health centers or geriatric homes, where patients with neurological conditions need more constant care or monitoring, supported by the YOLO Object Detection algorithm in version 3 and the DeepSORT algorithm. For the follow-up of people, it is proposed to generate a program in Google Collaboratory using the Python language for the detection, monitoring and evaluation of physical activity of people through video surveillance or computer vision, with a real-time approach based on various AI techniques. These algorithms have various uses and advantages, some examples range from tracking the ball in football or basketball matches to detecting cars for automatic driving systems, among the advantages it offers is that it is a non-invasive, low-cost technology and High efficiency, on the other hand, presents some difficulties related to effects such as occlusion, that is, when one person is in front of another and makes it difficult to see them in the video. The large number of people can be another limitation, since in general a great computational capacity is required to detect each object, this intrinsically affects our performance in real time, limiting the number of frames that can be processed each second.

publication date

  • June 4, 2021 8:06 PM

keywords

  • Artificial intelligence
  • Deep learning
  • DeepSORT
  • Physiotherapy
  • YOLOv3

Document Id

  • fac35c91-db68-4a7c-8184-62e32a7cdfe7