Machine vision based method for automated defect detection of raw silk
Conference: MEMAT 2022 - 2nd International Conference on Mechanical Engineering, Intelligent Manufacturing and Automation Technology
01/07/2022 - 01/09/2022 at Guilin, China
Proceedings: MEMAT 2022
Pages: 4Language: englishTyp: PDF
Authors:
Xu, Shaolei; Huang, Yang (School of Mechanical and Electrical Engineering, Guilin University of Electronic Technology, Guilin, China)
Ge, Guoping (State Key Laboratory of Raw Silk and Silk Products Testing, Technical Center of Nanning Customs District, Nanning, China)
Liang, Yu (College of Life Sciences, Guangxi Normal University, Guilin, China)
Abstract:
Defect and cleanness detection plays an essential role in evaluating the quality of raw-silk grade, which affects the selling price greatly in the market. However, the grading of raw-silk is still mainly relied on the manual inspection to identify and detect the rough defects. To improve the efficiency and accuracy in raw-silk detection, this paper proposed an automated machine-vision based inspection method by analysing the texture and characteristics of raw-silk in the captured online images. The image of raw-silk is firstly segmented according to the texture by image binarization with a proposed two-step local threshold segmentation. Then the specially defined row and column vectors are implemented to remove the raw-silk backbone in each segmented image. Finally, the defect detection of raw-silk is achieved by morphological operation and features identification. The proposed method is numerically verified and experimentally validated in a developed setup, conforming its feasibility in automated raw-silk inspection.