Chinese Journal of Lasers, Vol. 47, Issue 1, 0114001 (2020)
Application of Wavelet Denoising in Terahertz Nondestructive Detection
Zhang Jiyang1,2, Ren Jiaojiao1,2, Chen Sihong1,2, Li Lijuan1,2,*, and Zhao Changshuang3
- 1Key Laboratory of Photoelectric Measurement and Control and Optical Information Transmission Technology,Ministry of Education, College of Optoelectronic Engineering, Changchun University of Science and Technology,Changchun, Jilin 130022, China
- 2National Experimental Teaching and Demonstration Center of Optoelectronic Engineering, College of Optoelectronic Engineering, Changchun University of Science and Technology, Changchun, Jilin 130022, China
- 393367 troops of the Chinese People's Liberation Army, Siping, Jilin 136000, China
This paper proposes a tomographic short-time integral imaging method. In terms of terahertz tomography, the experimental results show that the data quality and imaging effect of the proposed short-time integral imaging method are better than those of the traditional method. In the wavelet denoising theory, a δ-σ evaluation rule is proposed based on the characteristics of the terahertz signal, and the optimal wavelet denoising combination (e.g., the sym7 wavelet with a decomposition scale of 5) is selected using the evaluation rule. Based on this, the short-time integral imaging experiment of nondestructive detection tomography of phenolic plastic samples is set up, and different wavelet denoising combinations are compared. The effect of wavelet denoising is compared from two subjective evaluation indexes of defect number and defect recognition rate and the objective evaluation index of Weber contrast. Results prove that the sym7 wavelet (with the decomposition scale of 5, soft-threshold processing) is effective in wavelet denoising of nondestructive detection signals of phenolic wedge defects. The background noise in the nondestructive detection image of the sample after signal preprocessing is effectively suppressed. The contrast effect between prefabricated defects and the background area is more obvious, and the internal structural changes in the sample can be detected easily and accurately.