An Unsupervised Neural Network for Loop Detection in Underwater Visual SLAM

Thispaper presents a Neural Network aimed at robust and fast visual loop detection in underwater environments. The proposal is based on an autoencoder architecture, in which the decoder part is being replaced by three fully connected layers. In order to help the proposed network to learn the features that define loop closings, two different global image descriptors to be targeted during training are proposed. Also, a method allowing unsupervised training is presented. The experiments, performed in coastal areas of Mallorca (Spain), show the validity of our proposal and compares it to previously existing methods, based on pre-engineered and learned descriptors.