研究目的
To propose and implement a sparse reconstruction-based technique for thermal imaging defect detection, comparing it with the cross correlation method in terms of defect detectability, SNR variation with defect depth, and computation complexity.
研究成果
Sparse reconstruction is an effective processing tool for thermal imaging defect detection, offering memory efficiency by compressing video data into a single image, higher SNR, and better resolution for close-spaced defects compared to cross correlation. It simplifies defect quantification and reduces computation time at optimal alpha values. Future work could extend to other materials and defect types.
研究不足
The study is limited to a specific CFRP sample with artificial defects; real-world defects may vary. The sparse reconstruction algorithm requires optimization of the user parameter alpha, which can be computationally intensive. The method's effectiveness for very small or complex defects is not fully explored. Thermal diffusion and noise may affect accuracy.
1：Experimental Design and Method Selection:
The study employs sparse reconstruction and cross correlation methods for processing thermal imaging data. Sparse reconstruction uses LASSO algorithm with l1-norm minimization and cross-validation for parameter optimization. Cross correlation involves offset removal and time-domain correlation with a reference signal.
2：Sample Selection and Data Sources:
A carbon fiber reinforced polymer (CFRP) test piece with artificially drilled cylindrical holes of diameters 4 mm and 6 mm, and depths from 0.25 to 2.5 mm, is used. Data is acquired using an infrared camera.
3：25 to 5 mm, is used. Data is acquired using an infrared camera.
List of Experimental Equipment and Materials:
3. List of Experimental Equipment and Materials: Includes an infrared camera (FLIR SC5000M), LED excitation source (40-W commercial LED), LDR for reference signal, microcontroller, comparator (LM323), potentiometer (10-k), and computer. The CFRP sample is 7 mm thick.
4：Experimental Procedures and Operational Workflow:
The setup involves modulating the LED with a quadratic frequency modulated signal (0.01 to 0.09 Hz for 800 s), recording thermal response with the IR camera at 1 fps for 1000 s, and processing the data offline using sparse reconstruction and cross correlation algorithms. Sparse reconstruction involves solving LASSO problems for each pixel with optimized alpha via tenfold cross-validation.
5：01 to 09 Hz for 800 s), recording thermal response with the IR camera at 1 fps for 1000 s, and processing the data offline using sparse reconstruction and cross correlation algorithms. Sparse reconstruction involves solving LASSO problems for each pixel with optimized alpha via tenfold cross-validation.
Data Analysis Methods:
5. Data Analysis Methods: SNR is calculated using Gaussian surface fitting for defect signal and RMS noise. Defect diameter estimation is done by fitting Gaussian surfaces to the thermal images. Computation time, sparsity, and l1-norm are analyzed for different alpha values.
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
