眼动追踪下产品外形特征提取算法设计与仿真  

Design and Simulation of Product Shape Feature Extraction Algorithm under Eye Movement Tracking

作　　者：林璐[1] 秦文斌 LIN Lu;QIN Wen-bin(Concord University College Fujian Normal University,Fuzhou Fujian 350117,China;College of Architecture and Planning,Fujian University of Technology,Fuzhou Fujian 350118,China)

机构地区：[1]福建师范大学协和学院,福建福州350117 [2]福建理工大学建筑与城乡规划学院,福建福州350118

出　　处：《计算机仿真》2024年第5期390-394,共5页Computer Simulation

摘　　要：不同产品的外形非常复杂且多样化,在产品外形特征提取过程中,由于受噪声、杂乱背景或其它干扰元素的影响,导致特征提取的准确性降低。为了精准提取产品外形特征,提出一种眼动追踪下产品外形特征提取算法。通过塔型方向滤波器分解带有噪声的产品外形图像,利用多尺度阈值对高频子带去噪,在有效保留信号系数的同时达到抑制噪声系数的目的,根据Contourlet反变换获取去噪后的图像。通过眼动追踪技术确定产品外形图像中人眼感兴趣的区域,在感兴趣区域内提取多个底层特征。使用改进的量子遗传算法对底层特征展开选择,最终实现产品外形特征提取。实验结果表明,所提算法可以有效提升产品外形特征提取结果的准确性和特征提取效率。The appearance of different products is very complex and diverse.In the process of extracting product appearance features,the accuracy of feature extraction is reduced due to the influence of noise,cluttered background,or other interference elements.In order to accurately extract these product appearance features,an algorithm under eye movement tracking was proposed.At first,the product appearance image with noise was decomposed by a pyramidal directional filter.Then,the multi-scale threshold was used to denoise the high-frequency sub-band,thus achieving the purpose of suppressing noise coefficients while effectively retaining signal coefficients.Based on the Contourlet inverse transformation,the denoised image was obtained.The area of interest in the product appearance image was determined by eye movement tracking technology.Meanwhile,multiple low-level features were extracted in the area of interest.Finally,the improved quantum genetic algorithm was adopted to select the low-level features.Thus,the product appearance feature extraction was achieved.Experimental results show that the proposed algorithm effectively improves the accuracy and efficiency of feature extraction.

关 键 词：眼动追踪 产品外形 特征提取 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]