Related Experiment Video
Updated: Jun 3, 2025
11:00
Experimental Investigation of Secondary Flow Structures Downstream of a Model Type IV Stent Failure in a 180° Curved Artery Test Section
Published on: July 19, 2016
11.6K
Focusing on Cracks with Instance Normalization Wavelet Layer.
Lei Guo1,2,3, Fengguang Xiong1,2,3, Yaming Cao1,2,3
1Shanxi Key Laboratory of Machine Vision and Virtual Reality, North University of China, Taiyuan 030051, China.
Sensors (Basel, Switzerland)
|January 11, 2025
Summary
This study introduces a novel Instance Normalization Wavelet (INW) layer to improve automatic crack detection. The INW layer enhances feature extraction and noise filtering, leading to better recognition and faster model convergence for crack segmentation.
Area of Science:
- Computer Vision
- Image Processing
- Deep Learning
Background:
- Automatic crack detection is crucial for structural health monitoring but faces challenges due to complex crack topologies, diversity, and background noise.
- Existing methods struggle with accurately segmenting thin, intricate crack patterns amidst noisy data.
Purpose of the Study:
- To develop an advanced deep learning layer for enhanced automatic crack detection and segmentation.
- To improve the accuracy, generalization capability, and convergence speed of crack detection models.
Main Methods:
- Introduction of a novel Instance Normalization Wavelet (INW) layer, inspired by wavelet theory.
- Embedding the INW layer into a deep learning model for crack segmentation tasks.
- Incorporation of a fusion layer to integrate multi-layer information.
Main Results:
- The INW layer effectively captures crack features while simultaneously filtering high-frequency noise.
- Instance normalization within the INW layer mitigates feature differences, improving model generalization.
- Experimental results on DeepCrack and CRACK500 datasets show steady enhancements in recognition and convergence performance.
Conclusions:
- The proposed INW layer significantly improves automatic crack detection and segmentation performance.
- The method offers a robust solution for challenging crack detection scenarios with complex features and noise.
- The INW layer accelerates model training convergence and boosts overall recognition accuracy.
