Objective   The defect detection of laser additive manufacturing (AM) has always been a technical problem that restricts its development. Due to the complexity of the defect generation mechanism, the insufficient detection information of the highly reflective workpiece surface, the low precision, the complex detection conditions, and other reasons, it is difficult to achieve high-precision and robust detection of defects. When the defect detection system based on reflective illumination performs detection on the surface of a metal part with high reflectivity, the pixels of the image sensor are usually overexposed due to the strong reflective light, resulting in a large amount of annihilated defect information, and it's difficult to highlight and extract the information of the defect area. Therefore, in view of the engineering problem of the high-precision robust detection and evaluation of surface defects of high reflective metal workpieces manufactured by laser AM, a polarization detection system based on a high reflective suppression effect is designed, which can effectively avoid background clutter interference and improve the defect detection capability in complex detection environment.

  Methods  The system is designed based on Q-type aspheric surface, which has a strong aberration correction ability and simplifies the system structure. The deviation between the seventh surface shape and the best-fitting spherical surface is only 0.371 μm (Fig.2-3). The deviation between the 9th surface shape and the best-fitting spherical surface is only 0.434 μm. The focal length is 50 mm, the number of F is 2, and the working distance is 300 mm.

  Results and Discussions   The simulation results show that the modulation transfer function is better than 0.42 at the Nyquist frequency of 144.93 lp/mm, meeting the requirements of the image quality (Fig.4). The tolerance analysis and 2 000 Monte Carlo analysis results indicate that the tolerance range is reasonable and meets the processing and assembly conditions under the condition of satisfying the image quality of the polarization detection system (Fig.10-11). To verify the suppression effect of the defect polarization detection optical system on the highly reflective light of the detection surface, the experimental device is built based on the designed polarization detection optical system (Fig.12). Based on the constructed polarization detection system, the detection images under different polarization angles are collected and converted from the RGB channel to the HSV channel for threshold determination. Furthermore, based on the Stokes vector method, the defect polarization information in the high-reflection detection image is extracted. The Stokes vector, degree of polarization, and angle of polarization detection image are calculated. The calculated image is fused to achieve the high-reflection suppression reconstruction of the defect detection image, thus achieving efficient and robust high reflection area characterization and analysis. The experimental results show that the fused image has a prominent role in the edge contour of the defect area, and the contrast between the defect area and the adjacent background area has been effectively improved, making the details of the defect clearer and more intuitive (Fig.14). The overall contrast, clarity, and information content of the image have been improved. Besides, to objectively and quantitatively evaluate the quality of the fused image and compare it with the original intensity image, the average gradient (AG), entropy (E), spatial frequency (SF), edge intensity (EI), and standard deviation (SD) are used to evaluate the image, the results are shown (Fig.15). Compared with the original intensity image, the average improvement rates of the average gradient, information entropy, spatial frequency, edge intensity, and standard deviation of the fused image are 163.46%, 20.04%, 163.20%, 123.03%, and 28.41% respectively.

  Conclusions   The results fully illustrate that the polarized image after fusion processing has more abundant information, the image details are clearer, the contrast of the defect area is higher, and the edge contour information of the defect is clearer. The feature extraction and characterization analysis of the surface defects of highly reflective metal workpieces in metal manufacturing are of great significance.