| The two-dimensionalization of traditional 3D textured surfaces is an urgent problem to be solved in computer vision,but most of the current research only focuses on the problem of 3D space or 2D planes,and there is a lack of relevant research on 2D 3D textured surfaces.Therefore,the two-dimensionalization of the traditional three-dimensional pattern surface faces great challenges.This paper takes the traditional three-dimensional pattern surface as the research object,and carries out the research on the two-dimensional method.This paper takes the lead in proposing a twodimensional algorithm based on the traditional process three-dimensional pattern surface,and builds a flattening system based on the traditional process surface pattern.The research contents and innovations mainly include:(1)A two-dimensional stitching method of three-dimensional surface patterns is proposed.The three-dimensional pattern surface is obtained by surface expansion and image stitching to obtain a complete twodimensional result.Firstly,it is necessary to obtain the 3D pattern surface segment through orthogonal projection,secondly,the non-developable surface is approximately processed into a developable surface,then the developable surface is expanded,and finally the multi-plane is spliced and fused.The above steps fully expand the three-dimensional pattern surface into a two-dimensional plane image.(2)A two-dimensional stitching algorithm for optimizing threedimensional surface patterns based on a single perspective is proposed.On the basis of the single-view stitching algorithm,taking the x-axis direction of the initial unwrapped image as the benchmark,and using the characteristics of the end-to-end connection of the three-dimensional craft surface rotating body,the stitching is continuously performed on both sides of the initial unwrapped image to reduce the deformation error of the image after stitching to ensure that the two-dimensional result after splicing is still a rectangle.(3)Subjective and objective evaluation indexes are proposed to measure the 2D result of 3D pattern surface.Aiming at the problem that there is no two-dimensional evaluation method for three-dimensional surface texture,subjective and objective indicators are proposed to measure the orthogonal projection segment and its two-dimensional results.The subjective indicators include:visual consistency,interaction consistency,and degree of difficulty.Objective indicators include cosine similarity and histogram similarity.(4)Developed a flattening system for surface patterns based on traditional craftsmanship.Using node.js and react framework,realize the flattening system based on traditional craft surface pattern.The system mainly includes a three-dimensional pattern surface two-dimensional module,a multi-angle projection segment two-dimensional module and a personal information module.The system embeds the trained algorithm model into the system background,builds a web front-end for the display of the input and results of the data source model,and enriches and optimizes user interaction.This paper proposes a two-dimensional splicing method for 3D surface patterns,which solves the problem of lack of related research on 3D surface patterns.The process solves the problem of slow and errorprone multi-image splicing;at the same time,it proposes subjective and objective evaluation indicators to measure the two-dimensional results of three-dimensional surface patterns,which solves the difficulty in evaluating the two-dimensionalization of three-dimensional surface patterns,which is usually observed with the naked eye.Finally,a flattening system based on traditional craft surface patterns is built to provide users with a friendly interface and help users realize one-click twodimensionalization according to their own needs. |