Research On Pattern Retrieval Algorithm Of Quasi-regular Pattern Based On Hand-sketching Contour

In recent years,with people's pursuit of personalized clothing,digital popular geometric pattern design methods based on theories such as chaos and fractal are favored by people because of their beauty and wonder,endless changes,high efficiency and convenience.As one of them,quasi-regular patterns have been widely used in garment pattern design and other fields in recent years.With the rapid popularity of flower patterns with quasi-regular patterns as the main design theme,their number also shows an exponential growth trend.The huge increase in number has brought about retrieval problems.It is of great significance in practical application that how to quickly and accurately retrieve patterns in a large library of quasi regular patterns and help people choose their favorite patternsThis article takes the quasi-regular pattern as the research object of the retrieval,based on the composition shape and symmetry characteristics of quasi-regular patterns,a related algorithm of quasi-regular patterns retrieval based on hand-sketching contour is proposed.The work of this article is mainly composed of the following three parts:(1)A quasi regular pattern retrieval algorithm based on color contour extraction is proposed.Firstly,the color region of the pattern in the quasi regular pattern is divided to extract the contour.Secondly,the shape feature and symmetry feature of the extracted color contour are extracted,and the similarity measure of the two features is calculated.Finally,the similarity measures of the two features are weighted and fused to realize the pattern retrieval of quasi regular pattern.The results show that the algorithm can accurately retrieve quasi regular pattern patterns with similar shape and symmetrical structure,and the average retrieval accuracy can reach 85.19%.(2)Based on the symmetry characteristics of the quasi-regular patterns,a quasi-regular pattern primitive segmentation method based on the iterative axis of symmetry is proposed.The method first uses the symmetry feature detection algorithm to obtain the symmetry axis of the quasi-regular pattern and the coordinates corresponding to each symmetry axis.Secondly,the initial position of the primitive segmentation and the length and width of the segmentation area are determined by the coordinates of the symmetry axis.Finally,the symmetric feature detection of the segmented patterns is performed continuously performed until the primitive patterns are segmented.Through many experiments,it is concluded that when the number of iterations of the symmetry axis is 3 times,the best primitive segmentation effect can be achieved.(3)A quasi-regular patterns retrieval algorithm based on hand-sketching contour is proposed.According to the symmetric structure of the quasi-regular patterns,the algorithm first uses the quasi-regular patterns primitive segmentation method based on the iterative axis of symmetry to segment the patterns in the quasi-regular pattern library.Secondly,in order to further highlight the main constituent shapes of the quasi-regular patterns,a preprocessing method combining the iterative optimal threshold segmentation method and the Canny edge detection algorithm is used for the primitive patterns.Finally,the HOG feature extraction algorithm is used to extract the shape features of the quasi-regular primitive patterns and the hand-sketching contours,and calculate the feature similarity measure of the two to realize the quasi-regular patterns retrieval based on the hand-sketching contour.The experimental results of the paper show that the mean average precision(MAP)of the retrieval algorithm proposed in this paper has reached 83.11%.