Extracting Object Heights From LiDAR & Aerial Imagery

TL;DR

The paper addresses how to extract object heights from LiDAR and aerial imagery using a reproducible procedural workflow. It combines SAM-based object segmentation with LiDAR-derived elevation maps to produce per-object height metrics, and contrasts this with a LiDAR-only, transformer-based encoding approach that fuses five bands. It discusses the current limitations of imagery-based segmentation and the performance trade-offs across resolutions, while outlining a future trajectory toward geospatial LLMs and GeoAI that could unify procedural methods with NLP. The work highlights practical, data-rich outputs suitable for geospatial research and sets a path for integrating remote-sensing data with transformer- and LLM-driven pipelines.

Abstract

This work shows a procedural method for extracting object heights from LiDAR and aerial imagery. We discuss how to get heights and the future of LiDAR and imagery processing. SOTA object segmentation allows us to take get object heights with no deep learning background. Engineers will be keeping track of world data across generations and reprocessing them. They will be using older procedural methods like this paper and newer ones discussed here. SOTA methods are going beyond analysis and into generative AI. We cover both a procedural methodology and the newer ones performed with language models. These include point cloud, imagery and text encoding allowing for spatially aware AI.

Figures (6)

• Figure 1: Point Clouds
• Figure 2: Results
• Figure 3: Procedural Method
• Figure 4: LiDAR Point Classification
• Figure 5: SAM Results