研究目的
To develop a non-destructive method for detecting debond defects in composite laminates using thermographic images, enhanced by Gabor filter and watershed segmentation techniques.
研究成果
The proposed method, combining watershed segmentation, active contours, and Gabor filtering, effectively detected some debond defects in composite laminates from thermographic images. However, not all defects were identified, indicating limitations due to image quality. The approach shows promise for non-destructive testing but requires further refinement, such as incorporating signal conditioning or additional imaging techniques, to achieve full defect detection.
研究不足
The method did not detect all defects present in the specimen due to poor signal-to-noise ratio and low contrast in thermal images, especially at image borders. Noise and artifacts from non-uniform heating and surface geometry affect detection. The study did not use advanced signal conditioning techniques like Thermographic Signal Reconstruction (TSR) or Principal Component Thermography (PCT), which could improve results. Future work may involve augmenting with other imaging modalities like guided wave ultrasound.
1：Experimental Design and Method Selection:
The study uses pulsed thermography to capture thermal images of a composite laminate with embedded defects. Image processing techniques including noise removal, histogram equalization, watershed segmentation, active contours, and Gabor filtering are applied to enhance and detect defects.
2：Sample Selection and Data Sources:
A composite laminate specimen of size 100 mm x 100 mm with 26 layers of CFRP at various angles (0/45/-45/90) and thickness of
3：4 mm is used. Teflon pieces (5x5 mm, 1 mm thick) are embedded at 16 known locations to simulate defects. List of Experimental Equipment and Materials:
Infrared Camera (Model Thermo CAM SC3000, 320X240 pixels, thermal sensitivity <20mK at 30°C), two flash lamps (Model Broncolor Topas A8, flash energy 6100 Joules each), composite laminate specimen, Teflon pieces.
4：Experimental Procedures and Operational Workflow:
Flash lamps are positioned at 45 degrees to the specimen at 50 cm distance. A heat pulse is applied, and the camera captures images every 10 milliseconds. Images are processed through steps: noise removal, histogram equalization, watershed segmentation (applied to whole image and divided quadrants), image fusion with different weights, active contours for defect sizing, and Gabor filtering for texture analysis.
5：Data Analysis Methods:
Defect detection is evaluated by counting the number of defects identified in processed images. Parameters for watershed (sigma and fraction) and Gabor filters (frequency, sigma) are varied to optimize results. Active contours are used to measure defect sizes.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
