Improving Post-Earthquake Crack Detection using Semi-Synthetic Generated Images

TL;DR

This work addresses the scarcity of labeled post-earthquake crack images by introducing a semi-synthetic augmentation pipeline that grafts computer-generated cracks onto photorealistic 3D models via parametric meta-annotations. The pipeline renders UAV-like scenes and produces corresponding annotations, enabling iterative tuning of augmentation to target detector blind spots. Evaluations on the IDEA dataset show that combining real and semi-synthetic images improves crack detection, with Many-to-Many metrics providing a more accurate performance signal. The approach offers scalable, controllable augmentation for mid-distance post-earthquake imagery and sets the stage for extending to additional damage types and video data.

Abstract

Following an earthquake, it is vital to quickly evaluate the safety of the impacted areas. Damage detection systems, powered by computer vision and deep learning, can assist experts in this endeavor. However, the lack of extensive, labeled datasets poses a challenge to the development of these systems. In this study, we introduce a technique for generating semi-synthetic images to be used as data augmentation during the training of a damage detection system. We specifically aim to generate images of cracks, which are a prevalent and indicative form of damage. The central concept is to employ parametric meta-annotations to guide the process of generating cracks on 3D models of real-word structures. The governing parameters of these meta-annotations can be adjusted iteratively to yield images that are optimally suited for improving detectors' performance. Comparative evaluations demonstrated that a crack detection system trained with a combination of real and semi-synthetic images outperforms a system trained on real images alone.

Figures (11)

• Figure 1: Workflow of the proposed semi-synthetic image generation procedure.
• Figure 2: Initial steps of the semi-synthetic image generation procedure: (a) Loading and examination of the 3D model; (b) selection of a relevant component to be damaged.
• Figure 3: Example of meta-annotations (black lines) and some details of two sample renderings showing cracks thus obtained.
• Figure 4: Example of the effect of the roughness parameter: (a) initial meta-annotation line; (b) low frequency distortion; (c) low and high frequency distortion.
• Figure 5: Different crack instances generated from the same meta-annotation (the red line).