Object Detection

The Object Detection tool allows users to identify buildings and vehicles contained in high resolution aerial imagery. The tool has two main components: Building Extraction, and Vehicle Detection.

This tool requires Global Mapper Pro.

This tool can be accessed from the Deep Learning (Beta) Drop-down Menu, or from the Deep Learning (Beta) Toolbar.

Building Extraction

Building Extraction allows for identification of building features from imagery. This process can create vector building footprints and a binary raster output. The model is trained on:

  • 3 spectral band aerial imagery, 0.3 meter resolution

Note: Any of these models can be fine-tuned as needed to improve outputs from similar sources.

Post-processing

Check the Apply post-processing option to modify the output building footprint area features:

  • Simplification - Check this option to simplify building footprints. This will remove extra vertices, resulting in a smoother appearance.

  • Regularization - Check this option to regularize building footprints so that corners become 90 degrees and opposite walls are parallel.

Output

Images above left to right: original image, binary raster output, building footprint area features.

Vehicle Detection

Vehicle Detection identifies vehicles in high resolution imagery. Vehicles are identified with a vector bounding box. The model is trained on:

  • 3 spectral band aerial and satellite imagery, 0.15 meter resolution

Vehicle Detection does not support imagery with a fourth alpha transparency band.

Output

The image above shows the vector bounding boxes generated identifying vehicles within the input imagery.

Custom Models

Select the custom models tab to specify a previously saved model created with the Fine-Tuning tool to use for building identification.