Clasificacion de la información tactil para la deteccion de personas

This paper presents the design of a tactile end-effector (EE) and the application of artificial intelligence techniques for human detection using a lightweight 6 DoFsmanipulator arm. This EE is composed of a high-resolution tactile sensor that allows to obtain pressure images. The system extracts haptic information in disaster situations where, generally, there is low visibility, in order to assess the state of the victims according to the urgency of care (triage). Two artificial intelligence methods have been implemented to classify images obtained by the haptic sensor, distinguishing contacts with people from inert objects in disaster scenarios. Each method has a pressure image feature extractor and a classifier, obtained by supervised learning. To validate the methods, classification experiments have been performed on Human and Non-human classes. Finally, a comparison of both methods in terms of classification accuracy and time has been performed based on the results of the experiments.