Abstract: In order to solve the deviation problem between tread pattern and design pattern in the process of tire vulcanization, a new method is proposed to extract and reconstruct tread pattern feature from tire tread scan data. Firstly, A new space segmentation method is determined according to the pattern pitch arrangement and the extrusion and calendering process, secondly, obtained the Contour of the tire crest by contour design theory and principal component analysis method, and finally successfully transformed the dimensional point cloud to the two-dimensional gray image (with depth value). According to the growth rhythm of the pattern and the idea of regional growth law, the pattern border characteristics can be identified and extracted accurately. The cubic boundary spline interpolation method is used to discretize the pattern boundary into a series of equidistant coordinate points. And finally, the approximation of tread curve of Self-adaptive Tangent Continuous Circle and Line is realized according to the definition of optimal uniform approximation. The experimental results show that the method can be used to reconstruct the tire tread pattern of various semi-steel radial tire in about 90s, the feature extraction error is less than 0.15mm, the mean value of reconstruction error absolute value is less than 0.25mm, the maximum value is less than 0.4mm. This method can extract and reconstruct the tread pattern boundary feature effectively and deal the distortion problem in pattern reconstruction of traditional reverse engineering technology, as well as provides a direction for improvement of the efficiency of pattern development and the comprehensive performance of the tire.

Key words: approximation, feature extraction, machining feature, reconstruction, reverse engineering, tread pattern