基于多尺度特征融合的道路裂缝缺陷检测方法及系统
By using a multi-scale feature fusion method, combined with infrared thermal imaging and 3D laser point cloud data, a 3D surface mesh model is constructed and curvature field analysis is performed. Features are extracted using a dual-branch encoder, which solves the problem of high false negative rate in crack detection under complex environments in existing technologies, and achieves high-precision and reliable crack detection.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- SICHUAN JISI DIGITAL INFORMATION TECH CO LTD
- Filing Date
- 2026-04-21
- Publication Date
- 2026-07-17
AI Technical Summary
Existing road crack detection methods struggle to simultaneously capture global structural information of both subtle early-stage microcracks and penetrating wide cracks in complex environments. Furthermore, the lack of explicit modeling in multi-source data fusion leads to a high rate of missed detections, threatening driving safety and structural durability.
A multi-scale feature fusion method is adopted. By constructing a three-dimensional surface mesh model with temperature attributes, and combining infrared thermal imaging and three-dimensional laser point cloud data, curvature field analysis is performed to extract regions with significant curvature. Multi-scale structural damage and surface continuity perturbation features are extracted using a dual-branch heterogeneous encoder. Iterative collaborative optimization is then performed to generate a crack probability map and perform geometric constraint screening.
It achieves high-precision detection of road cracks in complex environments, reduces the false detection rate, ensures the reliability and practicality of the detection results, and can adaptively restore crack information at different scales.
Smart Images
Figure CN122116148B_ABST