用于指尖定位的多目标分布估计算法-《郑州大学学报(工学版)》

文章信息/Info

Title:: A Multi-objective Estimation of Distribution Algorithm for the Fingertip Localization

摘要:: 在基于指尖的人机交互系统中，指尖中心的位置是非常重要的.通过求解指尖定位的多目标优化模型，能够得到多个指尖中心的位置.由于指尖像素点分布在指尖中心的附近，指尖定位优化模型的最优解分量也符合这一分布规律.求解指尖定位优化模型时，采用符合这一分布规律的分布估计算法，能够得到比较准确的结果.本文对指尖定位的分布估计算法进行研究，提出决策变量的维数、种群规模、最大采样方差、最小采样方差，是这一分布估计算法的主要参数. 实验结果表明：上述主要参数的取值存在最佳值；它们均取最佳值时，得到的指尖中心位置，优于已有方法的计算结果.

Abstract:: In the human-computer interaction system based on fingertip, the position of fingertip center is very important. By solving the multi-objective optimization model for the fingertip localization, several fingertip center positions can be obtained. The fingertip pixels distribute around the fingertip centers, so the optimal solution components of this optimization model have the above distribution law. Using the estimation of distribution algorithm with the distribution law to solve this optimization model, can obtain accurate results. This paper discusses the estimation of distribution algorithm for the fingertip localization. It proposes that the decision variable dimension, population size, maximum sampling variance, and minimum sampling variance are the main parameters of this estimation of distribution algorithm. The experimental results show that each main parameter has its best value; when their values are their best values, the fingertip center positions obtained by the proposed method excel the results of the existing methods.