基于遗传算法的DFD参数变形模型轮廓提取  

Footprint Boundary Finding Based on DFD Parametrically Deformable Models

作　　者：童莉[1] 平西建[1] 李存宝[2] 

机构地区：[1]信息工程大学信息工程学院,河南郑州450002 [2]信息工程大学科研部,河南郑州450002

出　　处：《信息工程大学学报》2007年第3期335-338,共4页Journal of Information Engineering University

基　　金：国家自然科学基金资助项目(60272004)

摘　　要：针对具有特性形状属性且轮廓局部光滑的图像目标,提出一种DFD(D iscrete Fourier Descriptors)参数变形模型轮廓提取算法。对于需要保留轮廓细节的轮廓形状,DFD参数轮廓变形模型需要对较大的参数集合进行优化,算法的速度和精度之间存在较大矛盾。本文利用遗传算法全局优化的性能和隐含的并行搜索特性,并通过初始种群选择约束目标整体形状属性,达到保留轮廓形状细节基础上提高遗传算法的有效性、加快模型优化速度的目的。将其应用于平面足迹重压区域轮廓提取,获得了较好的效果。For image objects constrained by a certain shape mode and smooth locally, this paper proposes a boundary finding method based on DFD （ Discrete Fourier Descriptors） parametrically deformable model. For extracting boundary with fine information, one must optimize the model on a larger set of parameters, which causes trouble in the balance of computing cost and accuracy. By applying genetic algorithms, a global optimal searching algorithm with better numerical stability and implied parallel computing mode, and introducing prior shape information in selecting initial group, this paper improves the performance of GAs as well as the speed of optimal searching. Better results are achieved in the application of this method on the boundary finding of weight bearing areas in footprint images.

关 键 词：图像分割 轮廓提取 参数变形模型 遗传算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]