Design of 3D LiDAR Combined with Neural Network for Object Classification

Abstract

LiDAR is one of the visual sensors which can measure a distance and form an environment description. This device is needed for many kinds of vehicle navigation especially for the autonomous system. Nowadays, the 3D LiDAR is still expensive in the market. This study has developed and constructed a 3D LiDAR consisting of a single point LiDAR as the main sensor and a Neural Network for classifying objects. Proportional-integral-derivative (PID) controller was involved to maintain the motor rotation in order to stabilize the scanning process. Arduino Mega microcontroller was used as the main processor to obtain the LiDAR data, to control the motor speed, and to communicate the data with computer. In this case, the 3D LiDAR was tested using five different objects. The experimental results show that the system can recognize all objects with a 100% success rate. This proposed system can be expected to support the road safety on an autonomous vehicle. In addition, the 3D LiDAR can be marketed in a low price.