手绘图符的辨识特征及其提取算法  

A Study on Hand-drawn Sketch Feature Recognition and Its Extraction Algorithm

作　　者：杨晨晖[1] 李剑[1] 刘聪[1] 

机构地区：[1]厦门大学信息科学与技术学院,福建厦门361005

出　　处：《厦门大学学报（自然科学版）》2014年第3期352-358,共7页Journal of Xiamen University：Natural Science

基　　金：福建省高校产学合作重大项目(2012H6024)

摘　　要：收集日常生活中常见目标的250类手绘图符20 000个,GB/T11797—2005标准中规定的28种机动车和非机动车的车符约2 400个,手写数字、手绘电路元件、手绘微软办公软件形状图各1 000个,作为本文研究的数据集基础.针对手绘车符不规则形变和样本间相似度高的特点提出一种基于局部细节和全局轮廓的特征表示方法,该方法利用弹性网格吸收手绘过程中的变形,利用局部梯度方向直方图(histograms of oriented gradients,HOG)表示图符的细节特征,利用Snake模型提取图符轮廓,傅里叶描述子提取低维轮廓特征,并通过融合使得整体效果提高,比单纯HOG算法高出2个百分点.We primarily use a sketch collection of common targets in daily life as the basic data set of the study.This data set in cludes 20 000 hand drawn within 250 classes,2 400 motored and non motorized hand drawn vehicle symbols of 28 kinds which are set forth in GB/T11797-2005 standards,1 000 handwritten numerals,1 000 hand drawn circuit elements and 1 000 hand drawn Microsoft Office software shape figures.To solve the problems of the irregular deformation of sketch vehicles and similarity among different vehicles during recognition,we propose a characteristic representation based on local details and global outline.This method is 2 percentage in recognition effects higher than the pure HOG algorithm by using elastic meshes to absorb the deformation of handdrawn,using local gradient histograms to represent the details,using the Snake model to extract the contour,using Fourier based descriptors to extract the low dimensional of the contour features,and making the overall effect improved by the fusion of these aforementioned methods.

关 键 词：手绘图符识别 弹性网格HOG特征 SNAKE模型 傅里叶算子 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]