For humans, recognizing personal protective equipment can be a natural ability. As for the virtual environment, it is not such an easy task. Machine language for object recognition involves a lot of calculations and requires a lot of processing power. However, with the evolution of the processing capacity of computers and evolution of algorithms, carrying out this activity is already possible. In this sense, the objective of this work was to build a personal protective equipment (PPE) recognition system
through machine learning techniques. In addition to doing an evaluative research with a quantitative approach through four major methodological steps in the research process to select a Machine Learning technique to be applied in the access control system. The developed project, the base structure is linked in the K-Nearest Neighbors techniques and the residual neural networks with ResNet-50 architecture. The techniques were constructed with the purpose of identifying the presence of PPE. Evaluations of techniques through statistical metrics and Machine Learning are presented. Finally, the simulation of access control is carried out for inspection of individual protection equipment.
|
