2020-08-142020-08-102020-08-142020-06-29https://ri.ucsal.br/handle/prefix/1665When a structure collapse occurs, caused by collapse or even for another reason, to obtain an optimized response time in the search and rescue of survivors, technology becomes an important ally in maintaining life. In this context, we see robotics as a process optimizer. This work deals with the construction of a Snake Robot, a system for the identification of living organisms in collapsed structures, which aims to identify living beings in these structures using machine learning, contemplating the implementation of an autonomous locomotion system, in addition to a classification of thermal images. For this, a thermal image dataset was generated, where we use data augmentation to add it in volume and variety, and from it generate a predictive model. This predictive model, in conjunction with object detection, allows the snake robot to function autonomously during the detection / classification of objects in living or non-living beings. In building the solution, sensors, an Arduino microcontroller were used, in addition to the Tensorflow library with Keras. With the simulated practice experiments carried out, the robot had its operation tested and validated, accurately detecting in a 180o field of view at distances between 10 to 150 cm, obtaining an average accuracy of 97.5 % for classification.Acesso AbertoRobô cobraMachine learningArduinoSnake robotTrabalho de Conclusão de CursoEngenhariasEngenharia de Software