Detect Damage by Images (Processing Pipeline)

Function Description: Integrates deep learning classification, image-based damage detection, PCI calculation, and report export pipeline, enabling one-click processing for image-based damage detection.

Input and Output

• Data Input

  • Lane Centerline: Different from the interactive selection mode in normal operations, here all vector lines within the "VirtualLaneCenterline" layer are used as input by default. Therefore, the lane centerline must be placed in the "Lane Centerline" layer in advance, while avoiding unrelated vector lines in this layer.

  • Point Cloud Data: Import all point clouds of the current project. The input point clouds can be unclassified, since a classification step is already integrated in the pipeline.

  • Image Data: If panoramic cameras are available, input all panoramic images. If no panoramic cameras are available but planar cameras exist, then input all planar images.

• Result Output

  • Report: By default, output to Current Project/Pavement Damage Analysis/DamageAnalysis.html

Steps

1.Click Processing Pipeline → Detect Damage by Images. The corresponding pipeline model will be automatically loaded, as shown below:

Detect Damage by Images Pipeline

2.If the pipeline fails to run due to incorrect input data or if custom parameter adjustment is needed, you may modify the input data or step parameters. Refer to Add and connect data and tools, and modify elements.

3.Ensure the pipeline is in a runnable state before proceeding. This pipeline supports distributed processing, with two execution modes: Run and Distributed Run (only Deep Learning Classification and Detect Damage by Images support distributed processing, other steps do not). For differences between non-distributed and distributed runs, refer to Run and stop a model.