Research on Digital Identification of Garment Pieces Based on Contour Fitting Algorithm

To help designers analyze the pattern shape of ethnic clothing and obtain its size parameters more conveniently, the contour fitting algorithm was proposed. This algorithm realizes the digital recognition and simulation of complex garment pieces, which combines image edge detection, Fourier descriptor reshaping, garment piece corner point detection, Bezier curve fitting and gradient descent. Compared with the traditional method, the garment piece corner point detection can effectively circumvent the noise generated by the threads of the garment piece. The gradient descent method improved the accuracy of contour fitting by adjusting the control points of Bezier curve. The accuracy of this method was compared the manual method for garment piece size measurement of Mongolian garments. The results show that the errors measured by the algorithm ranged from 0.06 to 0.18 cm, and their average error (0.12 xm) is close to the manual measured error (0.13 cm), This algorithm is in accordance with the GB/T 2660–2017 and can be applied to ethnic clothing restoration design.