Technology Category
Computer Vision: SegmentationIntelligent Robots: Multimodal Perception & Sensor FusionKnowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
LiCAR: pseudo-RGB LiDAR image for CAR segmentation
🤖 AI Summary
LiDAR point clouds are inherently incompatible with mainstream vision models designed for RGB inputs. Method: This paper proposes a pseudo-RGB spherical range image (SRI) representation tailored for vehicle instance segmentation—projecting LiDAR points onto a spherical surface and encoding reflectance, near-infrared, and signal intensity as native R, G, and B channels, enabling standard vision models without camera fusion. Contribution/Results: We introduce the first LiDAR-SRI benchmark dataset dedicated to vehicle segmentation and implement an end-to-end detector-segmenter based on YOLOv8-large. Experiments achieve 88.0% mAP@0.5 for vehicle detection and 81.5% Mask AP@0.5 for instance segmentation on SRI, while supporting robust multi-object tracking. This work is the first to empirically validate the efficacy of purely LiDAR-derived pseudo-RGB representations for fine-grained vehicle parsing, establishing a novel camera-free paradigm for 3D perception.
Technology Category
Application Category
📝 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.
Problem
Research questions and friction points this paper is trying to address.
LiDAR
Image Segmentation
Neural Networks
Innovation
Methods, ideas, or system contributions that make the work stand out.
LiCAR
YOLO-v8
LiDAR image processing
🔎 Similar Papers
No similar papers found.
