LiCAR: pseudo-RGB LiDAR image for CAR segmentation

TL;DR

This paper introduces LiCAR, a pseudo-RGB image derived from LiDAR data, and demonstrates its effectiveness for car segmentation using neural networks, achieving high accuracy and real-world tracking performance.

Contribution

The paper presents a novel pseudo-RGB LiDAR image dataset and applies instance segmentation neural networks for car detection, showing promising results.

Findings

Segmentation accuracy of 88% bounding box and 81.5% mask with YOLO-v8.

Effective tracking of segmented cars in video sequences.

New dataset combining LiDAR reflectivity, infrared, and intensity images.

Abstract

With the advancement of computing resources, an increasing number of Neural Networks (NNs) are appearing for image detection and segmentation appear. However, these methods usually accept as input a RGB 2D image. On the other side, Light Detection And Ranging (LiDAR) sensors with many layers provide images that are similar to those obtained from a traditional low resolution RGB camera. Following this principle, a new dataset for segmenting cars in pseudo-RGB images has been generated. This dataset combines the information given by the LiDAR sensor into a Spherical Range Image (SRI), concretely the reflectivity, near infrared and signal intensity 2D images. These images are then fed into instance segmentation NNs. These NNs segment the cars that appear in these images, having as result a Bounding Box (BB) and mask precision of 88% and 81.5% respectively with You Only Look Once (YOLO)-v8 large. By using this segmentation NN, some trackers have been applied so as to follow each car segmented instance along a video feed, having great performance in real world experiments.