Post Earthquake Wall Crack Mapping with Handheld Images and Mobile Depth Cues Sensor Data

Abstract

The rapid and accurate assessment of structural damage following seismic events is critical for ensuring public safety and facilitating timely recovery efforts. Traditional manual inspection of wall cracks is often hazardous, labor-intensive, and prone to subjective interpretation. While recent advancements in computer vision have enabled automated crack detection using conventional digital photography, these two-dimensional approaches fundamentally lack the spatial context necessary to accurately quantify crack width, depth, and structural severity in a physical three-dimensional space. This paper presents a comprehensive methodology for post-earthquake wall crack mapping utilizing handheld images combined with depth cues obtained from consumer-grade mobile devices equipped with Time-of-Flight sensors. By integrating high-resolution optical imagery with dense depth maps, the proposed system transcends traditional semantic segmentation to achieve physical dimension estimation and three-dimensional spatial mapping of structural fissures. The methodology encompasses multi-modal data synchronization, robust feature extraction through a specialized deep learning architecture described theoretically herein, and geometric projection techniques that translate pixel-level detections into real-world measurements. Extensive experimental evaluations demonstrate that the incorporation of depth data significantly mitigates false positives caused by surface anomalies, shadows, and debris common in post-disaster environments. The results indicate a substantial improvement in both detection reliability and dimensional measurement accuracy compared to purely vision-based baselines. This research bridges the gap between sophisticated structural health monitoring requirements and the accessibility of ubiquitous handheld technology, offering a scalable solution for decentralized disaster response and infrastructure triage.