AI-Integrated Building Modeling Solution

Solution Overview

SuperMap iDesktopX provides an AI-integrated building modeling solution that automatically generates city-scale 3D building models from oblique photography 3D model data. The solution involves the following functions: Generate DOM, binary classification, raster to vector, building footprint regularization, roof classification, height value extraction, and construct buildings with roofs. The workflow is illustrated below:

Workflow

Using the oblique photography 3D model shown below as an example, this section outlines the modeling workflow:

1. Generate DOM

This step converts oblique photography 3D model data into Digital Orthophoto Map (DOM) data in *.tiff format.

Implementation: Load oblique photography 3D model tile data into a 3D scene, then use the Generate DOM function. For optimal results, set the output resolution between 0.3-0.5 meters.
The resulting DOM data (coordinate system: 4326) is shown below:

2. Data Preparation

Using DOM data from Step 1, this phase performs binary classification, raster to vector conversion, building regularization, roof classification, and height extraction to produce standardized building vector polygons with height and roof type attributes.

  1. Extract building footprints: Use the Binary Classification tool with DOM data to obtain raster building footprints.
  2. Convert to vector data: Apply the Raster to Vector function to convert raster data into 2D vector regions.
    Resulting 2D building vector polygons (coordinate system: 4326):
  3. Regularize building footprints: Use the Regularize Building Footprint tool to remove invalid features (e.g., holes, redundant vertices). Manual optimization includes deleting non-building features, filtering small anomalies, and editing erroneous polygons.
    Regularized building vector data:
  4. Roof classification: The Roof Classification tool analyzes DOM data to identify roof types, storing results in vector polygon attributes.
    Unique value thematic map based on roof types:
  5. Extract building heights: Using the Extract Height Values tool with oblique photography data, height values are extracted and stored in vector polygons.
    Height visualization using Stretch Mode:

3. Generate Urban Building Models

This final phase constructs city-scale building models using prepared vector data. Both untextured white models and textured models can be generated.

  1. Data preparation: For datasets exceeding 10,000 features, use Rebuild Spatial Index to optimize performance.
  2. Construct white models: The Construct Buildings with Roofs tool batch-generates untextured building models.
  3. Generate textured models: Utilize the Procedural Modeling Toolkit (Access path: SuperMap Tech Hub -> Supplementary Resources -> 3D New Features -> 2024 -> AI+Procedural Modeling Toolkit) for batch production of textured building models.