Basic Concepts

Before starting with MipMapEngine SDK, understanding some core concepts will help you use it better. This chapter will introduce the fundamentals of 3D reconstruction in an easy-to-understand manner.

🌍 What is 3D Reconstruction?

3D reconstruction is the process of creating 3D models from 2D images. Imagine you've taken multiple photos around a building, 3D reconstruction technology can:

1. Analyze the relationships between these photos
2. Calculate the camera positions at the time of capture
3. Calculate 3D information for corresponding pixels based on camera positions and textures
4. Express the 3D geometric information of objects or scenes in the form of point clouds or meshes
5. Add textures to the model based on photo textures to build realistic 3D models

Application Scenarios of 3D Reconstruction

• Surveying and Mapping: Generate high-precision topographic maps and orthophotos
• Urban Planning: Create 3D city models for planning and design
• Cultural Heritage Protection: Digitally preserve historical buildings and artifacts
• Emergency Response: Quickly obtain 3D information of disaster sites
• Engineering Monitoring: Monitor construction progress and changes
• 3D Content Asset Generation: Generate 3D content assets for games, films, AR/VR

📸 Photogrammetry Basics

🔄 3D Reconstruction Process Explained

🔄 ReconstructFull Basic Processing Flow

Processing Flow Features

• 🚀 Fully Automated: Automatic processing from input to output without manual intervention
• 🎯 Intelligent Decision Making: Automatically select optimal processing parameters based on data characteristics
• 📦 Multi-format Output: Support simultaneous generation of multiple formats to meet different application needs
• ⚡ Parallel Optimization: Multiple output branches can be processed in parallel to improve efficiency

Output Selection Recommendations

• Web Display: Choose 3D Tiles + DOM Tiles
• Professional Analysis: Choose OSGB + GeoTIFF + LAS
• Universal Exchange: Choose OBJ + PLY
• Surveying Applications: Choose GeoTIFF + DSM + Control Point Optimization

1. Aerial Triangulation

AT is the first step in 3D reconstruction. Its tasks are:

• Calculate the precise position and orientation of the camera when each photo was taken
• Establish geometric relationships between photos
• Generate sparse point cloud structure of the scene

2. Dense Reconstruction

With camera positions established:

• Calculate depth for each pixel
• Generate dense 3D point cloud

3. 3D Model Reconstruction

• Build 3D mesh models from point clouds
• Create model textures from original images
• Generate LOD models for large-scale scene rendering

4. Different Format Output Generation

Finally, generate different outputs based on your needs:

• 3D Models: OSGB, 3D Tiles, PLY, OBJ, FBX and other formats
• Point Cloud Data: LAS, PLY formats
• Gaussian Splatting Data: PLY, Splats formats
• Orthophotos: Georeferenced images in GeoTIFF format
• Digital Surface Model (DSM): Terrain elevation data

Standard Output Directory Structure

All reconstruction tasks will generate the following standard output:

output/
├── 2D/
│   ├── dom_tiles/      # Orthophoto tiles
│   ├── dsm_tiles/      # Digital surface model tiles
│   └── geotiffs/       # GeoTIFF format results
├── 3D/
│   ├── model-b3dm/     # 3D Tiles model format
│   ├── model-osgb/     # OSGB model format
│   ├── model-ply/      # PLY model format
│   ├── model-obj/      # OBJ model format
│   ├── model-fbx/      # FBX model format
│   ├── point-ply/      # PLY point cloud format
│   ├── point-las/      # LAS point cloud format
│   ├── point-pnts/     # PNTS point cloud format
│   ├── point-gs-ply/   # PLY Gaussian Splatting format
│   └── point-gs-splats/# SPLATS Gaussian Splatting format
├── AT/
│   ├── mvs.xml         # AT results
│   └── mvs_undistort.xml # Undistorted AT results
├── report/
│   └── report.json     # Quality report
└── log.txt             # Processing log

Output Format Description

Format	Purpose	Features
3D Tiles	Web Display	Supports LOD, suitable for Cesium and other platforms
OSGB	Professional Software	OpenSceneGraph format, widely supported
OBJ	Universal Model	Simple and universal, easy to edit
LAS	Point Cloud Processing	Standard point cloud format, includes classification information
GeoTIFF	GIS Analysis	With geographic coordinates, can be used for measurement
Tiles	Online Maps	Multi-level slicing, fast loading

🎯 Key Parameter Explanation

Resolution Level

Controls the level of detail in reconstruction:

Level	Description	Use Case	Processing Time
1	Ultra-high precision, highest geometric detail and texture clarity	Professional surveying, fine modeling	Longer
2	High precision, some simplification of geometric detail, highest texture clarity	General applications, quick results	Medium
3	Low precision	Preview, quick validation	Shorter

Image Overlap

🔍 Quality Control

Factors Affecting Reconstruction Quality

1. Image Quality

   • Clarity (avoid blur)
   • Lighting conditions (uniform lighting is best)

2. Capture Parameters

   • Overlap (>70%)
   • Flight altitude (affects ground resolution)
   • Capture angle (vertical + oblique combination is best)

3. Typical Control-Free Accuracy

   • RTK/PPK: Centimeter-level accuracy (12cm + 12*GSD)
   • Regular GPS: Meter-level accuracy

💡 Best Practices for Reconstruction Accuracy

• 🔧 Reliable Reconstruction Accuracy: RTK and PPK control-free solutions can achieve good accuracy most of the time, but control points and check points remain the most reliable methods for accuracy assurance and verification. If your application requires 100% guarantee of accuracy targets, or project delivery requires sufficient evidence to support the accuracy of results, be sure to set up control points/check points, otherwise you may face rework of field data collection.

🚀 Next Steps

Now that you understand the core concepts, you can:

• Read the API documentation in detail and customize your best automated workflow based on your business needs