Desarrollo de un sistema basado en visión artificial para el reconocimiento y control del EPP en el personal operativo de la línea de extrusión de alimentos para mascotas petfood en balanceados EXIBAL matriz Chambo

Abstract

The aim of this work was to develop a system based on artificial vision for the recognition and control of PPE in the operating personnel of the pet food extrusion line "Petfood" in the Balanced "EXIBAL" Chambo matrix in order to lessen the harm caused by human negligence due to the non-use of PPE. The implemented methodology was based on experimental theoretical research following a series of ordered steps; The initial situation of the processes, workplaces, and protective equipment used by the operating personnel was determined by an on-site visit to the company. By means of the INSHT matrix, the most critical workplace was determined and this was established as the environment for the implementation of the prototype. For the design of the system, the need for indispensable components such as the image acquisition device, the Single Board Computer, the preventive and corrective signaling, as well as the complementary elements and the 3D printed protective case were determined. Python was used as a programming language, an NVIDIA Jetson Nano B01 as SBC, the Deep Learning ResNet 18-Body model was used for the segmentation of the human body, and the color detection was carried out with Artificial Vision. In this way, a prototype was developed that can detect the correct use of personal protective equipment with an efficiency of 72% and a response time of 7.38 minutes. The operators were also trained and a survey was applied to determine the degree of acceptance of the prototype, where a favorable 80% was obtained. This resulted in a satisfactory implementation of the prototype, which is an adjunct to controlling the correct use of PPE at the factory.