Get3D Knowledge Center

02 Content-Aware Functions

3 min read

Key Functions of Content-Aware Technology

Summary: Content-aware reconstruction uses AI to identify scene elements (vehicles, vegetation, buildings, water, thin structures) and apply category-specific processing strategies.

Who This Guide Is For: 3D production teams, photogrammetry operators, and quality managers dealing with ghost vehicles, exploded buildings, vegetation noise, and other common urban reconstruction artifacts.

Get3D’s Content-Aware Technology leverages AI to fundamentally transform how 3D reconstruction handles real-world complexity. Here are the key functions in detail.

1. Moving Vehicle Removal

The Problem

Standard photogrammetry assumes the scene is static. When a vehicle moves between exposures, it appears at different positions across images and triangulation fails.

Result: ghost geometry — smeared, semi-transparent vehicle shapes fused into the road surface, duplicated car silhouettes, disrupted texture.

How Get3D Addresses It

Stage 1 — AI Detection

  • Deep learning detection model identifies vehicle instances across all input images
  • Each detected vehicle is masked out of the images it appears in

Stage 2 — Multi-View Consistency Verification

  • Cross-validates detections across overlapping views
  • Preserves parked vehicles (static across all images)
  • Inpaints background road surface from adjacent clean views

Result: Clean road mesh with no ghost geometry — no manual correction required.

Vehicle removal key function

See deep dive: Moving Vehicle Removal — Full Analysis

2. Vegetation Mesh Simplification

Improves efficiency with lightweight processing of vegetation by reducing image usage and simplifying matching parameters, without affecting non-vegetation structures such as buildings.

Vegetation simplification

3. High-Rise Low-Overlap Optimization

The Problem

Oblique aerial photography achieves consistent ground sampling distance and image overlap at ground level. But when buildings rise significantly above the terrain, the effective overlap on vertical facades drops.

Consider a typical urban capture configuration:

  • Ground overlap: 80% lateral, 5 cm GSD
  • Rooftop overlap (for a tall building): as low as 40%, 1 cm GSD

The result of insufficient facade overlap is the well-known “exploded building” artifact: facades that appear shattered, fragmented, or extruded outward from the building footprint.

"Exploded building" artifact

How Get3D Addresses It

Use AT optimization to reduce ~80% manual modeling work and significantly improve production efficiency.

Result: During reconstruction, low-overlap high-rise areas are optimized and adjusted to mitigate “exploded building” artifacts, achieving a 90% optimization rate.

High-rise optimization result

4. Thin Structure & Hole Repair

The Problem

Thin structures are a fundamental challenge for photogrammetric reconstruction. Fences, signboards, utility poles, guardrails, and overhead cables are physically real features, but they are so narrow that they appear in only a fraction of overlapping images, and depth estimation across their surfaces is unreliable.

How Get3D Addresses It

Content-aware reconstruction handles thin structures through a multi-stage pipeline:

Thin structure pipeline

Key innovations in this pipeline:

  • Thin structure detection flags these elements before reconstruction
  • Adaptive weighted meshing adjusts triangulation to span gaps and produce coherent surfaces even with sparse point support
  • Independent front/back texture mapping ensures both sides of flat vertical structures are correctly textured

Thin structure result

Result: Thin structures represented as coherent, textured surfaces with both sides correctly resolved.

Thin structure final result

5. Feature-Limited Area Reconstruction

Feature-limited area reconstruction

6. Water Detection & Fill

Water detection and fill

Source: Get3D Knowledge Center

Graph View