Implementación de un sistema de visión artificial sobre Willow para recrear movimientos que ayuden en el aprendizaje y estimulación motriz de niños de 3-5 años de edad.

Abstract

The objective of the research was the implementation of an artificial vision system on Willow to recreate ovements helping in learning and motor stimulation in 3-5-year-old children. Information was collected from research carried out about motor stimulation in preschool-age children, from this information was possible to determine the therapy that was best suited to this research. During the development, the design and implementation of the prototype were carried out, which consists of two sections, the design of the structure of the humanoid Robot and the design of the artificial vision system. The algorithm was developed in Python furthermore, the OPEN CV library was used, to work the artificial vision system with the ARM architecture of the raspberry Pi4 card. The first step for the development of the algorithm is the acquisition and filtering of the image in real-time. Once the respective libraries have been loaded and all camera parameters have been calibrated, the segmentation and image processing are carried out. In this stage, the TensorFlow neural network was used, in the case of Willow_VA, the algorithm was programmed to detect the upper middle part of the body. Once the data package of the system is obtained, they are sent through serial communication to the Robot's microcontroller so that the movements can be executed. Through the validation tests carried out, it was possible to conclude that the prototype was well received by the teachers and children, as well as the prototype had an average efficiency of 90-100 in terms of percentage. It is recommended for future research to work with a paid deep learning code framework due to TensorFlow lite has certain compilation problems for communication and execution with higher complexity routines.