摘要
针对传统图案几何纹样提取方法无法得到最优特征向量,导致提取精度低、提取时间长、提取过程抗干扰性较差的问题,提出基于Bhattacharyya(简称B氏)距离准则的图案几何纹样提取方法。分析常见的图案几何纹样特征;利用Bhattacharyya距离准则算法,非线性映射图案几何纹样样本至高维核空间,获取最优特征向量,线性映射最优图案几何纹样样本至低维特征空间,并将核函数应用于映射过程中,提升算法全局收敛性,增加图案几何纹样特征与其它类别的Bhattacharyya距离,提高图案几何纹样提取精度。实验结果表明,采用该方法提取加入干扰噪声的图案几何纹样,提取的平均时间较短、提取提取精度较高,且提取得到的几何纹样清晰度较好,实验结果表明所设计方法的提取性能佳。
Traditionally,the extraction methods can not obtain the optimal feature vector,resulting in low extraction accuracy,long extraction time and poor anti-interference ability.Therefore,a method to extract geometric pattern based on Bhattacharyya distance criterion was proposed.The features of common geometric patterns were analyzed.Bhattacharyya distance criterion algorithm was adopted to nonlinearly map the samples of geometric patterns to high dimensional kernel space,so as to obtain the optimal feature vector.Meanwhile,this algorithm mapped the samples of optimal geometry patterns to low dimensional feature space linearly,and the kernel function was applied in the mapping process to improve the global convergence of algorithm.Thus,the features of geometrical patterns and other types of Bhattacharyya distance were increased to improve the extraction accuracy of geometric pattern.Experimental results show that the proposed method can extract the geometric pattern with interference noise in less time,and the extraction precision is higher.The extracted geometric pattern is clear.The experimental results prove that the proposed method has better extraction performance.
作者
周峰
班倩
ZHOU Feng;BAN Qian(Hubei University of Technology,Hubei Wuhan 430068,China)
出处
《计算机仿真》
北大核心
2021年第8期492-496,共5页
Computer Simulation
关键词
距离
图案
几何纹样
最优特征向量
Distance
Pattern
Geometric pattern
Optimal feature vector
作者简介
周峰(1970-),男(汉族),湖北应城人,硕士,教授,研究方向:包装设计、视觉传达设计、交互设计、ui设计;班倩(1994-),女(汉族),湖北荆州人,硕士生,研究方向:ui设计、包装设计、视觉传达设计、交互设计。
