Defining materials using laser signals from long distance via deep learning

Abstract

Material identification is useful in robotics, industrial manufacturing, autonomous driving and so on. Cameras are generally used in material identification studies. However, in cases where lighting conditions are not suitable, material detection with cameras is difficult. In the proposed system, the defining of the target material, is realized with the LSTM deep learning model using only one laser light source that works independently of the environment light. Objects located at a certain distance are marked with a low-powered laser light source and laser signals reflected from the objects are recorded with the sensor system. After the data preparation steps, laser signals are trained with the LSTM deep learning model and the classification process is performed. For this purpose, laser signals recorded from 10 different materials at a certain distance, which are frequently used in daily life, are accurately classified with an average of 93.63% with the LSTM model. Experimental studies show that material defining can be performed using the LSTM deep learning model from a single laser measurement point. (c) 2021 THE AUTHORS. Published by Elsevier BV on behalf of Faculty of Engineering, Ain Shams University. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/ by-nc-nd/4.0/).