A Pattern Adjustment Method Based On Objective Function Optimization

The manufacturing process of clothing for sewing products has three steps: designing a pattern(including adjusting the pattern);using the pattern to make a template;cutting the fabric according to the template and sewing.Among them,the pattern adjustment process performed on human bodies of different sizes is also called a grading process.A universal and effective pattern algorithm that does not require any prior experience can not only improve the realism of virtual humans of different sizes in the entertainment industry,but also provide powerful productivity for the pattern design process in the apparel manufacturing industry.In the existing grading technology,there are problems that the poor robustness causes the target human body that is significantly different from the basic human body cannot be adjusted,the highly nonlinear target function leads to slow optimization speed,and poor optimization stability causes convergence failure.In order to solve the above problems,this paper proposes an improved pattern adjustment algorithm based on the objective function.For the objective function,a new boundary shape factor is proposed in this paper.This factor includes the relative rotation control of the boundary which the original boundary shape factor missed,which solves the robustness problem that failed to convergence to a reasonable solution when the size of the target human body changes greatly from the basic human body.For the optimization algorithm,this paper proposes a new L-BFGS hybrid acceleration optimization method,which combines the advantages of the complete convergence properties of L-BFGS and the accelerated convergence properties of other methods,greatly improves the speed of optimization on the premise of ensuring convergence to the optimal solution.For the optimization process,this paper proposes a new triple approximation alternative strategy,which solves the problem that the gradient oscillates over time during the optimization process,causing the optimization process to slow down and sometimes even fail to converge,greatly improving the gradient backtracking phenomenon,effectively improve the stability of optimization,and ensure the rigor of physics.This paper considers a variety of different garments,performs universal grading experiments on a large number of common-sized human bodies,and performs robust grading experiments on exaggerated human bodies.Based on the analysis of the fitting degree and style reduction level of the grading result,this method accurately considers the local details of the human body,and the adaptation of the grading result to the target human body is very close to the basic situation,which proves that the method is correct,universe and robust.