识别手绘形状的两种算法的实现  

Two methods for hand-drawn shape recognition

作　　者：屠红蕾[1] 

机构地区：[1]北京师范大学珠海分校信息技术和软件工程学院,广东珠海519085

出　　处：《计算机工程与设计》2009年第10期2513-2515,2520,共4页Computer Engineering and Design

基　　金：澳门科技发展基金项目(018/2005A);北京师范大学珠海分校重点基金项目(Z06007)

摘　　要：研究并实现了人机交互中大量使用的手绘形状识别算法。对在线和脱机两种情况分别使用了两种算法。针对在线识别使用了连续线段傅立叶描述子,然后归一化,对归一化后的系数进行测试样本和训练样本的比较。对脱机识别使用主成分法,获得训练样本的主成分,将测试样本和训练样本在主成分空间上进行投影,就获得的投影分量进行比较。用跨平台的Java语言实现了算法和识别系统,对不同形状进行了实验并给出了实验结果,同时给出了两种算法特点的比较。Hand-drawn shapes are widely used in human machine interactions. Their recognition algorithms is researched. For the online recognition, the algorithm of normalized Fourier transform descriptor is used. Via computing the continuous Fourier transform descriptor of the lines of the drawing trace, matching is based on the normalized FFT descriptors of the test samples and the training samples. For the off-line recognition, the principal component analysis method is applied. By extracting the principal orthogonal subspace of the training samples, the recognition is accomplished by computing the samples＇ projections onto the subspace and then comparing them. The recognition system is implemented in Java, which is a platform independent language. The lab results are presented and the comparison of the two algorithms is provided.

关 键 词：形状识别 手绘形状 傅立叶描述子 主成分法 JAVA 跨平台 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]