Extract Building Footprint from Image
Description
This function uses deep learning to extract building footprint from images. The input file must be a DOM (digital orthophoto image) or a three-channel color image produced by aerial photography, such as an orthophoto image. It cannot be a single-channel image such as DEM or DSM. The output is a gpkg vector file containing multiple building footprints.
Usage
Click on 3D Building > Extract Building Footprint from Image.
Parameters
- Select file list：Select image data either from already opened software or from a folder. Currently supported formats include tif, tiff, jpg, jpeg, and png.
- Simplify（default“Yes”）：When simplification is checked, the building outline will be streamlined to some extent, eliminating the jagged edges in the original outline.
- Regularization（default“Yes”）：Whether to perform regularization on the generated building outline. After regularization, the angle between any two adjacent sides of the building outline will not exceed the angle threshold.
- Angle Threshold（default“10”）：Line segments with angles smaller than this threshold will be merged into a straight line.
- Delete（default“10”） ：If checked, contours with an area smaller than this value will be deleted.
- Output Path ：The path where the output contour vector file will be saved.
Input
High-resolution imagery (10-25 cm) with 8-bit depth and at least three bands.
Model Architecture
The model is built on the YOLO v8 architecture from the Ultralytics package.
Accuracy Metrics
The model achieves a mean Average Precision (mAP) of 77%.