Skip to main navigation Skip to search Skip to main content

Comprehensive Comparative Analysis of RandLANet and PointNet++ for Aerial LiDAR Semantic Segmentation

  • Shivaram Kumar Jagannathan
  • , V. Sravani
  • , Jhagruth Palakonda

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Semantic segmentation of aerial LiDAR point clouds underpins applications in urban planning, environmental monitoring, and disaster management. A thorough comparative study of two leading point-based architectures, RandLANet and PointNet++, has been conducted using the DALES dataset. Quantitative evaluation includes Global Accuracy, Mean Accuracy, Mean IoU, and per-class IoU, supported by an expansive literature review of 18 foundational and recent works, ablation studies on block partitioning and sampling strategies, and computational trade-off analysis. RandLANet attains 95.78% Global Accuracy and 0.702 Mean IoU, outperforming PointNet++ (93.65%, 0.534). These findings guide the selection of point-wise segmentation models for large-scale airborne LiDAR.

Original languageEnglish
Title of host publication2025 Control Instrumentation System Conference, CISCON 2025
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798331597733
DOIs
Publication statusPublished - 2025
Event2025 Control Instrumentation System Conference, CISCON 2025 - Hybrid, Bangalore, India
Duration: 01-08-202502-08-2025

Publication series

Name2025 Control Instrumentation System Conference, CISCON 2025

Conference

Conference2025 Control Instrumentation System Conference, CISCON 2025
Country/TerritoryIndia
CityHybrid, Bangalore
Period01-08-2502-08-25

UN SDGs

This output contributes to the following UN Sustainable Development Goals (SDGs)

  1. SDG 11 - Sustainable Cities and Communities
    SDG 11 Sustainable Cities and Communities

All Science Journal Classification (ASJC) codes

  • Control and Systems Engineering
  • Electrical and Electronic Engineering

Fingerprint

Dive into the research topics of 'Comprehensive Comparative Analysis of RandLANet and PointNet++ for Aerial LiDAR Semantic Segmentation'. Together they form a unique fingerprint.

Cite this