A method for intelligent detection of micro-cracks under texture interference of a drain pipe
By constructing a prior feature encoding layer and a decoupled feature fusion algorithm, the problem of accurate differentiation of microcracks in drainage pipelines under complex texture interference was solved, achieving lightweight and real-time detection, and improving detection accuracy and intelligent operation and maintenance.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- KUNMING UNIV OF SCI & TECH
- Filing Date
- 2026-04-10
- Publication Date
- 2026-07-17
AI Technical Summary
Existing microcrack detection technologies for drainage pipes struggle to accurately distinguish between the background texture and microcrack features under complex texture interference, resulting in high rates of missed and false detections. Furthermore, the detection models are highly complex, making it difficult to meet the requirements of lightweight design and real-time performance.
A prior feature encoding layer adapted to the inner wall features of drainage pipes is constructed. Combining the inherent texture and structural feature differences of the pipes, a feature decoupling module and a decoupling feature fusion algorithm are inserted through the YOLOv8 infrastructure to optimize the bounding box loss function and generate a lightweight microcrack-specific detection network.
Significantly improves the accuracy and location capability of microcrack detection against complex texture backgrounds, enabling real-time end-side detection of microcracks on the inner wall of pipelines, reducing dependence on external computing resources, and improving the level of intelligence in operation and maintenance.
Smart Images
Figure CN122415469A_ABST